Detect Fraud with Power BI And Gen AI | Fraud Detection Using Power BI And Gen AI | Simplilearn

Right. Okay. Uh I can see a few responses so please uh share your names as well. Hi um hi from Shillong. Uh welcome to the webinar. It's wonderful to have you with us. Anybody else joining us from different cities? If you're joining us from uh any other part of the world, please let us know. Hi Prai from Gazyabad. It's wonderful to have you. I'm also getting a few responses from YouTube and LinkedIn. So we have uh Cavi from US. Welcome to the webinar. We have Rachel from US again. Uh welcome to the webinar. Thank you for joining us. Hello. Hello Shrea from Maharashtra. Hello Rhiti. Hi Mohammad from India. Hi Praachi. Praachi. I I saw your message again. Uh anybody else joining us? Please tell us which city you're joining us from. Hi Anil. Uh welcome to the webinar. Thank you for joining us. So many introductions coming in. Uh so I I don't think I'll be able to take up all of it. Uh please share the YouTube link as well. You can go to the uh you know uh YouTube channel or simply learn and you can join uh from there. All right. Uh so it's time for us to start the webinar. So once again welcome everybody. I'm so glad all of you could make it today. And if you have ever wondered how banks catch fraud in real time or how you as a business analyst or data professional can build that kind of intelligence yourself, you're exactly in the right place. Today's session is called how to catch credit card fraud using PowerBI and generative AI. And let me tell you, we are not just talking theory today. We are going to walk through a real project demo built live with real data using tools you can pick and use uh tomorrow. And let me also clarify that this is going to be a project demo webinar and not a live workshop. So you will not be able to go do this along with sol. So we request all of you to pay attention to the webinar. The recording will be shared with all of you. If you have any doubts, you can definitely ask us in the Q&A box and we will be very happy to answer that. We will try to keep the short as well. We want to show you what you can learn through our courses and also how to do this as well. And uh finally I am I I would also like to introduce myself. I'm your host I'm Schwa and I'm joining from Simply Learn site and we have a wonderful speaker join us today as well. I'll give his introduction very shortly. Um and I have a quick question for everyone in the chat. Please let me know. Are you uh currently working with PowerBI in any way or are you just getting started? Are you completely new to this? Because uh you know your experiences matter to us. Please let us know before we get into the webinar. New to PowerBI. A lot of people who are new here. A very good place to start. We will address that as well. Uh so uh thank you for all of your responses. Thanks Shia Widgets Nitish. Everyone thank you for sharing your responses. All right. So before we begin a few quick ground rules for everyone. uh if you have any questions I request all of you to share your questions in the Q&A box and not in the chat because uh of the uh high number of participants we have there there are chances that we might miss out on your messages. Um and we also request you to keep the conversations related to today's uh topic. Don't share any external links in the chat and that would be distracting and I see a lot of questions uh about the certificate of participation. So after this webinar we will be launching a poll towards the end. you can uh respond in the poll and you will be receiving your certificate slide tech and today's recording uh in your email ID within 48 working hours. So I hope that addresses all of your questions and with that we I want to quickly uh take 2 minutes of your time and tell you about simply learn. Let me introduce simply learn to some of you who are joining us for the first time. So at simply learn what we do uh everything revolves around just one idea and we would love to help professionals build skills that actually move their careers forward. That's what we've been doing over the years and we have worked with millions of learners worldwide across ROS industries and career stages and we have more than 50 partnerships globally. Speaking of partnerships, here are some key partners that we work with. The programs that we develop and provide are in association with prestigious institutions like UCSD extended studies per Michigan engineering professional education as well as leading global part organizations such as Google, Microsoft and IBM. Uh I also wanted to point out some of the key highlights uh of our courses. So most of our learners have reported a 50% salary hike on completion of courses with us and we have also been rated 4.8 8 out of five by most of our learners which clearly talks about the kind of learning experience you can experience with us at simply learn through our webinars or through courses. We we are very glad that you could join us today. So uh now uh coming to the most important part for me um and something that I really enjoy doing is introducing our guest speaker. So uh today we are joined by Solomon Promise. He is an engineer uh AI engineer, machine learning instructor and an edtech founder who has mentored over a thousand learners globally in data analysis, data science, machine learning and also AI. He has also led real world AI projects including heading a global team to build an AI powered traffic system and today he is actively shaping how people learn and apply AI through his work in education. And what I personally find exciting is that Solomon has a unique ability to break down complex concepts and turn them into something practical, something that you can apply. So, uh, as we go into today's demo, I'd highly encourage all of you to really pay attention not to what he's doing, but how he thinks through the problem because that's exactly what we want you to understand because sessions like this are a rare chance to see how experts approach real world data step by step. and we really hope you have a wonderful session today with Solomon. So Solomon, once again, we're really excited to have you with us. So over to you. I'm sure everyone's waiting to hear more from you. Hi everyone. I hope you can all hear me. It's it's a pleasure being here with you and I'm hoping that today's session would be an insightful session. Yeah. Thank you so much for having me. over to you. Yes. Uh thank you so much. I'm sure they're all waiting to hear more from you. So I'll dive right into the webinar. We have a lot to cover. So let's talk about the problem we are here to solve today. And I want to frame this clearly because sometimes people think fraud detection is is just a big bank problem, something for massive financial institutions with hundreds of engineers. It actually is not. It affects every organization that processes payments. uh and every analyst who works in finance as well. So and there are everyone that's affected includes compliance teams even risk managers as a matter of fact and the current systems are sometimes they are failing and uh we have the uh statistics for that the stakes that are mentioned. So uh first there are some staggering numbers out there on the screen. So first consider this there's $33 billion is lost every year to credit card fraud globally. So this has a global impact for businesses and consumers alike. And the second uh point that we have is that the project the projected annual cost is a jaw-dropping 10.5 trillion. So this is a huge number. To put this into perspective, this amount surpasses the GDP of every country in the world except the US and China. So that's the kind of financial threat cyber crime poses at a global scale today. And finally more than half 50 7% of the fraud cases involve digital transactions as per the ACF report 2024. So as we move forward we'll explore why the dete traditional methods are failing and the opportunities for smarter solutions as well. So uh let's move into the main part of it and Solomon before we move ahead I think I have a very uh important question to ask you. We've talked about fraud increasing, but what's really going wrong behind the scenes with how companies are trying to detect fraud cases today. All right. Thank you. So I I think one of the things still going wrong is that um some companies still try to detect fraud using what we call rulebased approach. Yeah. So rulebased approach is a kind of approach whereby you have to configure or come up with a certain rules and say if a transaction happens like this then it's a fraudulent transaction. If it doesn't happen like this then it's a fraudulent transaction. The thing around that is that most of these fraudulent guys or individuals they get to understand those rules and eventually they adapt to it. Yeah. So rulebased or traditional rulebased approach fails. Yeah. So that's why it's very important to always integrate machine learning. Yeah. AI into how fraud is being detected. Yeah. So AI is able to spot patterns that rule base won't be able to um spot and most of the times it hides in plain sight that you won't be able to see and some other times not even in plain sight it hides in a very large amount of data set. Yeah. So having machine learning in your fraud detection system or having a form of intelligence that's not just rule based in your fraud detection system sets you just ahead of these fraudulent persons. Yeah. And some analysts are stuck to the old way of doing things. They are very conversant with using Excel. They also very conversant with using maybe some PowerBI. Yeah. But they really do not have some knowledge about AI and ML. Yeah. So incorporating AI and ML to your skill set as an analyst equips you much better to help detect all of these problem cases. Yeah. Thank you. Over to you. Yes. Uh thank you. Uh and I think that what you mentioned is a very important point today. how much generative AI can help you or assist you uh you know uncover all of this and help us surpass these issues that we have facing currently. So with that I would also like to point out that um there is a great opportunity out there. So instead of reacting after fraud happens you could actually start identifying patterns before it escalates like you mentioned with generative AI and that is what we want to show you in today's uh webinar as well. Uh so uh again a major question to all of you what if you can uh fix it with a fraud detection workflow which you are going to see a live demo of. Uh so first we today we're going to look at real transaction data because everything starts with an understanding of what's actually happening. Then we bring that into powerbi and build a fraud monitoring dashboard. This is where patterns start becoming visible and from there we'll create visualizations that help us spot anomalies things that wouldn't be very obvious in with raw data and this is where it also gets really interesting. We'll use some amount of generative AI to actually explain those patterns turning data into clear business uh insights and by the end of this you won't just see a dashboard you will understand the skill sets needed to do this in a real world role. Uh so with that I think um I'm going to have uh to ask Solomon this question. So if someone is seeing this for the first time, it's a lot of information. How would you break down this workflow step by step to someone for them to quickly understand the concept? Because I understand a few of them are completely new to this. So could you please help them with that? All right. So there is usually a workflow for any data analytics tool you are working with. Yeah, there's a workflow and this workflow usually starts with a raw data set. Yeah, I'll be showing you a raw data set today. So, if you're a data analyst, you'd always you definitely would have worked with a data set before now. And if you are just coming into the world of data analysis or machine learning, data science, you definitely start up with a data set. Every problem starts with a data set. However, before that data set, there's always a problem. So the actual workflow starts with your problem definition which is basically saying what am I trying to solve? In our case we're trying to detect fraud. Now that fraud can be fraud within your organization. Yeah it can be different kind of fraud. Maybe not necessarily credit card fraud. Yeah it could be I think I once worked on a project some years ago which was detecting fraudulent health care providers. So it could be fraudulent healthare providers. Yeah. And so on and so forth. But the place you start is with being able to um really really put your problem into proper proper perspective to say this is the problem we are actually trying to solve. Now that determines the next course of action which is your data set. So the problem you're trying to solve determines the data set required to solve that problem. So you have to go gather your data set. If you are working within an organization your data set may be available there. Yeah. If you're working in maybe a banking system or a financial sector and so on and so forth, you definitely have the data set of all of those financial records. Yeah. So after that, in this case, I'll be using a synthetic data set. Yeah. Because transa financial transactions are very very important data set. Yeah. Personalized and so on. So you don't just get to put them out there. So but in a real world scenario within your organization, you definitely be working with yeah real data set from your organization where you have transaction ID, amounts, merchant categories, a whole lot of information embedded into that transaction um data set and most times this data set is historical. Now you have to gather a lot of historical data sets. Yeah. Because if you're going to integrate some level of AI Yeah. and some level of ML. Yeah. So what we're talking about here is not necessarily ML. By ML I mean machine learning. Yeah. But one of the ways to actually really really detect fraud is by using machine learning systems. Yeah. That are that can be predictive or that are predictive systems. Yeah. So one of the things you usually do is that your historical data sets do not usually have the label that this is fraudulent, this is non fraudulent. So someone would have to do that manually or an AI would have to do that. Yeah, some yeah some generative AI can help you read through the pattern and identify the product transaction or maybe you can pass it the pattern to say this is the pattern and that's not the pattern and so on and so forth. But basically you start out with having a very huge raw transaction data. Now if you work with a reward data set your data set would always be messy. Yeah. Most of the times yeah it would be messy and this is where you have to clean your data sets and structure your data set. It can take enormous amount of time cleaning your data set. And this is because your data set determines the output you get. Yeah. Whether you're building a Powerboard, PowerBI dashboard or you're building a machine learning predictive model or whatsoever you are doing the data set you use. Your results is largely dependent on your data set. So you have to do a whole lot of data cleaning, making sure your data set is ready, checking for things like data types, missing values and so on and so forth. In PowerBI, we have some more specific things like calculated measures and calculated colons. Maybe I'll show you that when we get into the demo session. Yeah, I so you can get a context of what calculated measures and calculated colons are. Now after that if you are using a machine learning system what we are doing there now is going to be purely on PowerBI. So on PowerBI what you now do is you have to build your visual visualize pattern detection. So this is where you build your visuals visuals that can help you um highlight highlight patterns. When we say product patterns the the core idea of prodent patterns are basically patterns that are not normal. Yeah. For example, if my transaction is always between $5 to $15 and all of a sudden there is a spike of $1,500 and it's happening for three consecutive time just within short periods of time then that can be already flagged here. So visuals, one of the things visuals do is that they give they give you insight into your data set. I'm going to show you something. Maybe I should just share my screen a little and show you something. So I'm asking for your request and let me share my screen. All right. Thank you. So take a look at this data set. This is the data set I would be using for the demo. I just want to ask you a question. Now looking at this data set, it's not a very large data set. Of course, this is the amount colon. So, I want to ask everybody a question and I'm going to give an award to anybody that gets the answer to my question. It's a very simple answer. It's not It's not tricky. Yeah, very simple and kind of straight to the point. So, look at that amount. Colon, I'm going to scroll down slowly. I have 50,000 rows. Now, I'm going to slow down. I'm going to slow down. I'm going to slow down. I'm going to slow down. I'm going to slow down. So if I ask you the question which is which of the transaction is the highest transaction? Yeah, you would if I give you one second to give me the answer, you'll be able to give me the answer in one second. Yeah, if I give you one second, yes, if you're proficient in Excel, you might be able to use some of the calculations. But if I show you a visual, if I show you a visual, once you see a visual, you definitely would be able to see what's going on. If I show you the visual and ask you a question like saying which of the category here has the highest spending amount immediately you can spot the category because it's a visual. So that's why or why visuals are very very important. Yeah they basically can allow you spot patterns that the human eyes can't spot in just a few minutes. Yeah. Thank you. You can share your screen again. Let me stop sharing mine. Yes. Sure. Sure. All right. So after um so after that you can now in use some generative AI layers. One of the good thing in today's world is that gen AI is being integrated into most of the tools we use. Yeah. For example in PowerBI you have copilot added to it. Yeah. I think you can also embed claude AI and some other gen AI tools in Excel. You can embed Claude AI directly to your Excel directly to your PowerPoint and so on and so forth. Yeah. So this makes it now it makes it easier because these gen AI tools can get into your data. They can get into the analytics paring your visual and you can ask them question. If you ever converse with Chad DPT or Cloud AI, you definitely get what I'm actually talking about here. So, you're able to ask your GI questions and it's able to give you insight. Yeah. It's able to also give insight to anybody that gets to use the dashboard eventually. Yeah. So, at and at the end of the day, you have a wonderful dashboard that can that stakeholders can now visualize interactive AI embedded. But it doesn't stop there. You can go on to build a fraud detection system. Yeah. Using powered by a machine learning predictive model. So most times fraud detection systems are a whole like it's it's a complete system and within that system you now have a machine learning model. For example, every one of us here has an email account. Maybe you use Google that's Gmail or you use a Yahoo mail or hotmail or any other mail you get to use on your mailbox there is a spam folder now that spam folder not just the spam folder anyway all all other folders are powered by ML models yeah so that's how when a message comes to your email box um it's being sent directly to an inbox or promotion or update then spam messages are sent to spam And if you had time to go through your spam, you would notice that some of the messages sent to your spam are actually spam. Yeah, they actually stars. Some of them will say you've won certain amount of money and so on and so forth. So basically I'm saying that yes, PowerBI is a very wonderful tool. It allows you visualize but beyond that you can also build predictive models that can be embedded into your system into the fraud detection system. Yeah. But you need a skill set to do that. I think we'll talk about that as we make progress. Over to you. Thank you. Yes. Uh thank you for providing such a detailed explanation Solomon. This really helps everyone understand the workflow better and they also can have an idea of what they can expect. There is also a very important question I'd like to address something that we received from the chat and on which we also have the upcoming slide. So uh the question is why is PowerBI such a powerful tool for detecting plot patterns and why are we doing this? Is it just for the visualizations or how how useful is PowerBI for someone who's new to this? Sorry, I was speaking and muted. Yeah. So why? Yeah. Can you hear me? Someone said I should increase the volume. Okay. Okay. So why PowerBI a very very powerful tool for fraud detection? So if you got the last thing I said I talked about PowerBI and I also talked about machine learning predictions or predictive models. So PowerBI has some very wonderful abilities. Yeah. One of the fact that is it allows real time data fer. What do we mean by that? So what we mean by that is I think I can share my screen again. So take a look at this dashboard. Yeah, take a look at this dashboard. This dashboard was built with a static data set. It was built with a static data set. But if I plug in this dashboard to a database, if it's connected to a database or if it's if maybe not even a database, just the Excel Excel Excel workbook and I have data set coming in at intervals. I have data set coming in at intervals. One of the ability of PowerBI is that it allows you do real time analytics. It allows you do real time analytics. So if I have a dashboard, I can because my data set is a synthetic data set. It it tries to capture real world scenario but it's not doing it excellently. So if you have a reward scenario, please share your screen. There's a slide I want to show them from your from your slide. Yes, I think that will help me help me portress my point. Thank you. All right, go to the next slide. The next one. Yeah. So let's take a look at this. Now if you take a look at this slide you would see that so I want to explain I'm explaining the like I'm responding to the question why PowerBI is a is an important tool in detecting prodent transaction and one of the things I've mentioned is that PowerBI allows for real time analytics real time analytics and the concept of real time is that as you have data coming in you can already begin to see your data on your dashboard board so you can already see what the data looks like and that's why I brought us here. So if this is my data set, if this is my dashboard rather, now as data is coming in, as data is coming in, I can already see data sets that are anomalies. I can already see data sets or data points. By data points, I'm referring to transaction transaction. I can already see transactions that are anomalies. Now that's one thing. But if I have a predictive model also embedded, one of the things the predictive model does is that it can already predict. Yeah. So it's able to flag that this transaction is a fraudulent transaction. So that's why they can when when a financial institution notice that this transaction is a problem transaction if they able to spot it immediately they're able to maybe put a hold on your account so that transa problem transactions do not take place and so on and so forth. Yeah, but one of the importance of PowerBI is that it allows for that real time analytics. That's one of the things. Yeah, you can go back to the previous slide. Yeah. Now the other thing it allows is what we call drill through capability. This is something you're all familiar with how by you can slice, you can drill down to see what is happening granularly. Now why machine learning predictions are very very powerful to they are they can predict that this transaction is going to be fraudulent. Yeah. What PowerBI does is that it allows you drill down. So you can see categories for example. So um you can see categories, you can see locations, you can see time within the day like the time bound band within the day where more transactions are likely to be flent based on historic pattern. Yeah. So because of that you already have insight. Yeah. So even if you're going to build a predictive model, you have an insight already to what the fraudulent um pattern looks like. Yeah. you have an insight to what the fra pattern looks like. It can again help you properly prepare your data set for a predictive model. Yeah. So I'm mentioning predictive model even if it's not on this slide because it's very very important. Now the other thing is in today's world where we have gen AI where we have gen AI I really can't remember exactly if PowerBI can flag but because we have gen AI tools in PowerBI they also give us an extra layer of insight. Yeah they also give us an extra layer of insight into the data. So if you ask me, it's going to be more of a collaboration between BI tools and predictive models. So I'm going to first build a a very detailed dashboard with my historical data set. Understand what the product pattern and so on looks like within the data set. Then after that I will go on to build a predictive model. Yeah. Because I already have a detailed understanding. Yeah. So for me PowerBI is very important at the beginning giving you a very detailed overview of what is going on within that landscape then you can move on. Yeah. Thank you. Over to you. Yeah. Uh so thank you for that uh Solomon. So with that can we dive right into the work the demo project demo? You can start with the demo and then we can talk about how to connect the data and build the foundation from there. All right. So, um this is a demo. It's a very short demo because we are like half way past our time. Yeah. Yeah. I'm going to stop sharing my screen. So, yeah. All right. Thank you. So, we're half we're halfway past the time of the entire work demo session for today. So what I'm going to do because a lot of persons are new to PowerBI when the facilitator asks I noticed that some of us are very new to PowerBI. Yeah. So I wouldn't already start teaching you how to detect stuff. I'm going to just give you a a slightly detailed overview into PowerBI so you understand what PowerBI is actually about. Then we can go on to do one or two more things. So take a look at PowerBI. This is what it looks like. Yeah. And you can build dashboards like this. You could build multiple dashboards. Yeah. Multiple dashboards. You could build multiple dashboard. But this is what PowerBI looks like. Now PowerBI has four views. It has four views. I'm going to slightly explain those views. Okay. All right. So it has four views. The first view is what we called report view. That's the view you can see at the moment. It's called the report view. That's where you build your visual. If I create a new page, you see that it's blank. It's a white canvas blank. So you build your visual here. That's the report view. Now the next view is called the data view. The data view is where you see your data. So take a look at this. You don't see anything here. It's basically a blank sheet. Can you guys hear me? All right. So it's it's basically a blank sheet. You can't see anything. But once I come to the data view or the table view, I can see my data set in a table. Yeah. I can see my data set in a table. So you see your data set. I actually have two data sets here. One of them is just a um a calendar. Yeah, one of them is a calendar. The other is my transaction data. It's usually good to have a calendar or a date table. Some persons we call it a date table. Basically, what it comprises are dates. Yeah, I autogenerated this one. That's why you can see 1950. Yeah. So, it's autogenerated. You can generate it to be custom fit to a particular date range. But I won't be doing that in this class. So, that's and why do I need to have a different date table? One of the important is that we do what is called data model where we connect the data set. So if I come to this you see that I have this transaction and this calendar. Let me show you something. So this is the calendar data and this is the transaction data. When I come to the third view which is model view I have both tables. I have both tables and I can connect them together which is what I want to do now. So I connect them together. It's called creating a relationship. So you're able to create a relationship between both tables. What it allows you do is that it allows you filter. Yeah, it allows you slice. It allows you filter. So this is not like a PowerBI class pass. So yeah, I won't be able to do all of those detail stuff, but just giving you an insight. So this is the third view called the model view. So we build data model. Basically data model is you connecting several data sets together. You can have four data set. You can have five data set and you need to connect all of the data set to together. It's called a data model. Yeah. If you are conversant with SQL, it's also something we call relational table or relational data table where you want to connect all the tables because they are related. Yeah. In PowerBI, one of the connections or relationships that are so so prominent and very useful, I'll say it so you can go check it out, is called a star schema. It's called the star schema. Star. Yeah, star schema relationship. It's a star schema relationship whereby you have one of the tables in the middle and you have the other tables connecting to it like a star. It's not always having the shape of the star. The core idea is that you have a table called a transaction table. You have a table called a transaction table. Just one table is called a transaction table. Then you have several tables called pack table. The difference is that your transaction table usually has numerical values. So if for for a typical let me use something we are very conversant with like the e-commerce. Yeah. Or a even even a credit card transaction. Yeah. So for a credit card transaction I already have a table called a transaction table. This calendar is a fact table. Packed table just contains categorical information. So I can have calendar. I can have demography. So all of the demography of my customers will be in one table. I can also have um demography here. I can have different payment method payment method all of the different payment method in a different table. I can also have something like different merchants or so on and so forth in a different table. And yeah but my center table is a fact table. It contains the detail. Yeah. transaction amount, items bought, items sold, all of those things that we can calculate, we can aggregate, they're usually in the dimension or in the fact table rather, sorry, they're usually in the transaction table. So, so that's that. Now, that's how we build a star schema. Why am I even doing all of this? I'm doing all of this because if you don't have any knowledge of PowerBI, then you can build a PowerBI dashboard. Yeah. And it would every other thing would make so much sense to you. Now we have a port view that is called the DAX view. DAX query view. DAX is simply data analytics expression is called the DAX view. And the last of them is T M DL. TMDL view. can't remember the full name but it's a more advanced view. It's used for scripting. It's used for scripting. In this class or in this session rather we'll be stuck with one two I'd use this. Yeah. One two three I've already connected this. Yeah. Then would also do this but not necessarily um I won't be I'll be writing dark query but I won't be writing them here. I'll be writing them somewhere else. So those are the three different very important views. Yeah, I don't have time to explain this. So we won't be doing that. On the reports panel, we have three pans to the right. We have the filter pan, we have the visualization pan, and we have the data pan. The data pan basically contains all of your data set. So what we can see here in a table we can see it here just listing out the columns listing out the columns alphabetically A B in in an alphabetical order. So these are the data then this is the visualization pan where we will build our visual from and this the filter where we'll be filtering from. Now I've built a few visuals. We do not have time to build from the scratch again. So I'm going to just show you what my visuals look like. Yeah. So 1 minute. All right. So I think you can still see my screen. Okay. Okay. Let me just show you this the way it is and we'll explore it further. Now take a look at my data. Click outside. All right. So take a look at my dashboard. Before I even show you my dashboard, take a look at the data because that's usually the starting point. Yes, this is my data set. I have a few column. The first column I have is my transaction. Let me increase my screen a little. Close all of this. Yeah, this is by the way, but you can see cloud here. Yeah, that's because I have a paid Claude account. So you could integrate all of these things. Let me hide this so it becomes more. All right. So we have transaction dates and transaction time. We have merchants. We have category. We have amount. We have amount. We have city, we have state, we have latitude, we have longitude, we have city population, job cons, job date of birth, transaction number, merchant latitude, merchant longitude, then is fraudulent. This is the colon I told you that you have to manually introduce yourself. I'll be very fast because of my time. I'm looking at my time. Yeah, our time is fast spent. All right, so take a look at my dashboard. What I intended to do was to build two different dashboard. One of them would show you what a normal transaction looks like and the other will show you what a fraudulent transaction looks like. But I don't have all of the time. So what I did was that I included a filter that can help you filter fraudulent transaction and legitimate transactions. But now take a look at the metrics. We have 50,000 transactions. We have a total of 9.8 million naira $7.89 million rather and an average transaction of 157.74. Yeah, $157.74. So that's the average amount spent. Yeah, this is for the entire data set regardless of whether it's fraudulent or non fraudulent. Now if we click on the legit transaction, 49% of the transactions were actually legitimate transaction. 49% of the transaction. So remember initially we had 7.8 but now we have 6.45 million. It means that over um over a million naira went past fraudent transaction. Yeah. So this business was losing a lot to fraudulent transaction or their customers that were losing a lot of fraudulent transaction. Yeah. Now if we come to let's focus on fraudulent because fraud is what we want to see. But just take a look at this. Yeah. Before we move in, these are for the cities. The different cities we have for the cities. Yeah. Any of the city you click on, you just see. Yeah. These are the cities. Their balls are almost the same because the amounts spent are similar. If the amounts spent are very different, in real life situation, the amount spent will be very very different. Yeah. But this is like I said a synthetic data set. Now this is what the transaction per day looks like. This is what the transaction per day looks like. There looks like there's a spike. Yeah. So these are the days of the month. That days of the month the number of transaction. Yeah. You can see the total transaction. The least transaction is on the 31st day of the month. I think the maximum transaction is on the sixth day. Yeah. Maybe they just paid salary there. So people had a lot of money to do some shopping. And maybe at this point, you can see that from the 30th to the 31st, even from this point, it start dropping. Yeah. Continuously. But now let's click on fraud, there are 1,00ent cases. Now we can see it. This was one of the things I was mentioning. Visuals already allow you see the problem cases. There are 1,000 fraudulent cases. Now the other thing that is very obvious take a look at my screen. The other thing that is very obvious is that most prevalent cases happened around online retail. Online retail the list was around wire transfer. The list was around wire transfer. Yeah. Online retail followed by electronics. So he's basically saying that within this my data set people are likely to um online transactions are more likely to or online retail are more likely to be fraudulent transactions. Yeah, we had a total of 1.44 1.44 million as fraudulent transaction. That's a lot of money going into fraudulent transaction. It's something to pay attention to. Then you can see what it looks like. There is a spike on the sixth day. 46 of the transactions are actually fraudulent. Let me show you what this implied. On the sixth day, on the sixth day, there is a total of 121 transactions. 100 1,721 transactions on the sixth day. Yeah. And out of that 47, sorry, let me do it again. There's a total of 1,721 transactions and out of that 40 36 are prevalent. 36 are problem. We can check the rates. Let me include that to check the the rate. What's the percentage of fraudent transaction? Where is my rate? All right. So, if I come here and make it big. Oh, no. That's That's not 100%. Yeah, give me a minute. Let me do something. Okay, I'm having an issue with that. I guess it's my an issue with my data. Yeah, I will not be able to fix it on the call. I'm so sorry. So what I intended you to see was the rate of fraudulent transaction actually have that then yeah so this was not the dashboard I intended to use but I just want you to see so entirely in my data set there are 2% product transaction 2% yeah so 2% accounts for the 1,000 product transaction we see What I actually wanted you to see was the percent of transaction within each of the days. Yeah. And it's showing me 100. Okay, I get why it's showing me 100 because I am selecting fraud and that's why it's showing me 100. So let me unselect fraud and let's come to the day six. Yeah. So day six there are 2% 2% fraud. If I come to day this Yeah. So can see this is 3% fraud. So out of the transaction the total transaction here 3% of that were actually fraudulent transactions. So you see why fraudent transactions are kind of dicey. They dicey because they are very take a look at this day 13 and the rate of fraud transaction is just one. Now this one transaction can be a lot of money. Yeah it can be a lot of money. Yeah, you can see the total fraud amount is $32,000. Just one fraudulent transaction. In this case, just two $55,000. Just two. Yeah. So, it can be very imbalanced. Just one transaction, but that one transaction can cost the client a lot of money if not dealt with. Yeah. So, we don't have time to drill down. So one of the things I wanted you to see is that you can drill down to the different states. Yeah. To see what the program transaction is like. It's completely different as we go from state to state. It's completely different. So you see that this is for let's go for Florida. Yeah. On the sixth day. Let me just stick with the sixth day. This is the total percent 2% but this is the amount. The amount of fraudulent transaction is just about 9,000 something. Yeah. Some states there might be no fraudulent transaction. Yeah. And you can see the total transactions in Florida is about 9,000 something. This is the total amount spent in Florida. If we check for the fraudulent, you can see in fraudulent in Florida 186 transactions that happened were fraudulent transaction. if we include a moon slicer um for a month. So I'll be wrapping up here so we can close within a time frame. Apologies. All right. So let me just show you this and we will wrap up. All right. So, so this is the total number of transactions that happened in Florida 971. Sorry, the total transaction is 9 9,957. The number of fraudulent transaction 886. Now we can begin to go January, how many transactions were fraudulent? February how many transactions you have problem March how many transactions. So we can already begin to see in the particular state for a particular state for a particular month the number of transactions that are problem the amount they amounted to also the total number of amount the average amount yeah so so you get basically so it allows us it allows us understand what's actually happening here within that I have a paid account but it's it's not accessible at the moment I would have showed you some of the capability of what copilot can do but you can't access copilot without a paid account. Yeah, you need a paid account to access co pilot. I think I would hand over to our facilitator for now. Yeah, I hope you learned a few things from this brief session. Over to Sure. So all right um I hope my screen is visible to all of you right now. Um there there were a lot of points that we wanted to cover a little more but again this was a project demo and we didn't have this was not a live workshop so Solomon could not take you through every step but however there are some questions about data sets we will try to share the data set that Solomon has used I'll check with him on that and uh now I I think we also have very limited time so we will not be able to cover all of the slides but we'll talk about some very important slides that we don't want to miss. And with that, I think Solomon, the next question I have for you is basically the different types of data. You already covered what normal data looks like. What does fraud data look like generally and I also I'm asking this question because from the audience point of view as well. While you were giving the demo, there were some questions about from Johnson who said how asked how how do they how to identify this kind of fraud and uh uh there's also someone else who had the same uh query about how this PowerBI is analyzing fraud. How does fraud in data look like and how do they analyze the same? Uh you are on mute so I don't think everyone's able to hear you. All right. All right. I hope you cannot hear me. Apologies. So like I said initially um fraudulent transactions are usually outliers. Fraudulent transactions are usually outliers. So when we are building visual we are more like doing fraudulent we are more like doing um anomaly detection that's more like what we doing we are trying to do anomaly detection. So when you are building visual what you try to do is that you want to build visuals that can actually spot or capture the anomalies. So it can be transaction amount anomaly it can be transaction frequency. It can even be transaction count. Yeah. So normally a person spends like he he uses his credit card like twice or thrice in a week and all of a sudden there are several debits on that credit card. So those are a few things to majorly holistically. So look at anomalies. Yeah it can be anomalies in transaction amounts. It can be anomalies in geographical impossibility. So for example, I am based within um Florida and someone is all of my transaction in the last 6 months has been within Florida and all of a sudden my card is being used from California or it's being used from Nigeria or it's being used from um a country in Europe or also Yeah. So it can be flagged. Yeah. It doesn't necessarily have to be but it can be flagged because it's an anomaly. It can also be off activity. Yeah. Maybe my activities are usually clustered within a particular time. The whole concept is the fact that anomalies are usually I'm more likely to be problem transactions. Yeah. More like Yeah. So I hope you get that. Over to Yes. Uh thank you so much for that. uh so I will also move on uh to another very important aspect aspect of today's webinar and that is the impact of AI on the role of a business analyst we do know that a lot has evolved we have been trying to say that with uh AI analysis has gotten better detecting these frauds have become faster so could you talk to us about the impact of gen uh generative AI when it enters the workflow does it help detect these faster and if so How do they do that? Yeah. So, like I showed you um on my Excel and I also showed you on PowerBI, most of the gen AI tools are now being integrated into our basic analytics work tools like PowerBI and Excel. Yeah. So PowerBI and um generative AI tools can actually now help you go beyond just the surface anomies, right? Can help you go beyond the surface anomies. It can help you flag outliers. Yeah. It can help you flag outliers. It can also even tell you why possibly. Yeah. Depending on your prompt. Yeah. So I think I I want to say this way. So we use we most of us are very conversant. I think just a few people in the world today are not conversant with at least charge. Yeah. So we use all of those things. We've seen the capability of those things in our normal day life. Yeah. So when we bring it into the context of fraud detection. Yeah. They can do a whole lot of things. Yeah. Helping us spot anomalies. Yeah. Helping us answer like the why question. Why is it even an anomaly? Yeah. Helping us even write reports. Yeah. So, a whole lot of things. I'm building a solution currently and it's a PowerBI solution, but we are working to integrate AI into it so that people are not just looking at dashboards. They are able to get other insights using generative AI. So, there a lot. It can actually do a lot. Yes. Um, thank you for saying that. And now we clearly understand the impact of generative AI in business analysis. Uh could you also help us quickly how to structure prompts effectively for uh fraud insights if that is something that they are looking for. All right. Yeah. So um prompting is very very important if you are going to use generative AI whether for frauding site or what have you. prompting is very important. I had the opportunity of picking some kids or things that like some weekends ago and and I gave them three key things that are very important for every prompt and it's something you can always remember. Yeah. One of it is the context. One of it is the context. The other one is the person. Yeah. And the other one is the question or the ask. Yeah. So let me explain that to you. So you need to give it a context. Yeah. So every time you use generative AI, you need to give it sufficient context. On this slide you can see on my screen you can see it's called branded context. You need to give it sufficient context. Now you need to also give it um you need to give it your act. Yeah. So like what do you want it to actually do for you? You need to be specified. Yeah. You also need to give it yeah your personal. Sometimes you need to tell the generative AI to something like you are a fraud detective um personnel or what have you. So basically you want to give it a personal that it's going to act like yeah those are a few things. Now this is beyond the concept the um domain of fraud detection whether you're using from gen AI for anything you need to give it option context ask it specific question yeah and give it a person yeah so that way it's able to sit like it's able to enter the costume of that your personnel and it's able to answer your question yeah in this case you need to do some other things like audience specification telling the AI model the compliance team basically giving it one of the things about Genai is that context is very very important yeah so giving it sufficient context it's very very important yeah I think that would go a long way to help thank you yes um thank you for clearing that out so you can definitely we'll be sharing the slide deck with all of you so you guys can take a look at the prompts that are there and how to structure it effectively and please do try this out as And finally uh Solomon I have one final question uh to you about how the role of uh uh you know a business analyst is now evolving with AI and how it how important is it for someone in the field or who's hoping to stay in the field to upskill themselves with uh generative AI with a course that has generative uh you are on mute again. The generative AI the analytics landscape is changing. Yeah. So we are moving from just traditional analyst to AI augmented business analyst. The same thing goes for models. So we are moving from models that can just predict to models that can just that can not just only predict, they can also generate. And we're also moving away from generative AI to models that are autonomous in nature. Yeah. So that's the same thing applies to a traditional analyst. Now a traditional analyst what PowerBI can do is that it can explain what's going on. So it describes what your data show. It's a static um report. Yeah. So a traditional ML a traditional business analyst would wait for the ML team to build fraud models. If you remember I mentioned the need of building ML fraud models. Yeah. If a traditional analyst, you actually can't build a fraud fraudulent model. Yeah. But when you argument AI as a business analyst, you are able to explain why the anomalies occur and what to do about them. So gen AI is able to go behind. Genai is able to go behind just telling you the what is happening. So in analytics there is the what is happening and there is the why is it happening. Yeah. There's the what is happening and there's the why is it happening. So your PowerBI dashboard and every other BI tool will tell you what is happening. Yeah. But Gen AI will go ahead to tell you why it's happening. Yeah. It's going to tell you why it's happening. Yeah. You can also build self-service dashboard. Like I said we are building a dashboard. Yeah. for a particular domain and we want to integrate AI in different layers of AI into it so that people are not just looking at static dashboards you're able to get insight ask it questions now because this AI are ML model powered so when we ask it question it goes beyond what I have built already it goes into my data set and tries to answer my question from my data. And the thing is with data is that your human eyes are limited to what you can get out of your data. Yeah. So, Gen AI helps you get out a whole lot of information that you can't get out of your data just using static model. How many if you look at my my dashboard, you'll also not like I think I had just about maximum of like 10 charts. That's not all that's happening in my data. That's not all. But with Gen AI, I can ask more questions. I can get more insight and so on and so forth. Yeah, I think that will be all. Thank you. Yes. Yes. Um, thank you. Thank you so much, Solomon. And, uh, that gives us a very good idea. I can also see a lot of questions in the Q. Uh, we'll try to take a few of your questions in the Q& uh, Q&A section which we have towards the end. Uh Solomon, if you find any questions in the Q&A box that you feel you can answer in the chat um or in the uh Q&A box, please do uh answer their questions. Uh if there's anything pending, we will try to cover that also. So with that, I think I'm going to move on uh to the next important part which is the uh AI powered business analyst course uh that we are offering at uh simply learn. And like I mentioned, this was a project demo webinar. This is something that you would be learning as a part of this course as well. And if what you have seen today resonates, if you're thinking I want to be able to build something like this, then this course would be something that could help you. This is an IIBA endorsed program aligned with guide week three. Uh you can earn 35 professional development units with CBAP focused live training. Uh it's also uses a AI approach here. That is what differentiates this program from others. So you will work with charg co-pilot and generated vi tools as a part of the core curriculum and this is not just an optional add-on. This is the core aspect that we are covering. You will also work on 10 plus industry projects and 40 plus case based activities with real world data sets and you will also finally get hands-on with the full tool stack. So there will be PowerBI, Excel, SQL, Tableau and a lot more and it is uh obviously mentorled live classes with uh live project guided project support. So these are not going to be pre-recorded videos that you are going to watch alone. You can ask your queries get everything answered as a part of this class. So that is what what we offer with uh this program that is the AI powered business analyst course and we're also very confident about the kind of course we provide. You can definitely go check out what this offer and also compare you to other courses if you may like. Our modules are very in-depth and that also brings you to uh to the learning path. So uh here on the screen I'm not go depth because all the modules are covered deeply in the course page. You can check out each each and every aspect to understand what what will be covered. So the core modules give you a complete foundation. So there's CBA preparation, AI powered analytics with Excel, generative AI powered SQL, PowerBI with AI assistance, RPM business automation and product thinking with agile uh frameworks. So if you notice something, the credit card fraud analytics case study uh that you just watched us uh create is also part of the curriculum and uh this will be a project you'll be working on and the electives also let you get a little more deeper. Uh so basically there will be Tableau certification, master generative AI, Microsoft as your data fundamentals. So these are electives as well. So this is not just a course. So we have a career architecture for all of you here. And uh when it comes to the uh skills and tools you can just look at the other the major tools like PowerBI, MySQL, Jira, Tableau, Lucid Chart, all of these are covered and the skills which are very important in today's market which includes business analysis, data elicitation, data visualization, data modeling, process mining, agile methodologies, design thinking, strategy analysis and more. So I'm not reading out everything because again the seeds are in the course page and I understand that we've gone a little beyond the time that we usually have for webinars. So I'm going to take you through this real quick. Uh this is a toolkit that you will require for an interview and this is what is going to differentiate you from all the others who are going to apply for the same right and these are some of the case studies and projects that are going to be covered as a part of it. One of it was credit card fraud analytics that you just saw. Again, this was not a workshop. So, you were not taken through it, but when you are a part of the classes, you're going to build this live. Some of the other examples of what you will be building are you're going to build a sales performance dashboard. You will be mapping employee performance metrics, data modeling again, and data visualization using PowerBI. Uh, you're muted. Uh sorry everyone. I'll just uh quickly share my screen. Um I had some issue on my end. Please give me a minute. I'll just fix this issue. Yeah. Uh so I hope uh my screen is visible to all of you. All right. So yeah, thank you all for bearing with me. I'll just take you through the skills and I'll wrap this up real quick. All right. Uh so like I was saying these are some of the skills and tools that we are offering as a part of this program. Uh and uh the case studies again I mentioned that this is there in the website. So you can definitely go ahead and check this out as well. Uh some of the other highlights and one of the most common questions that we get is once I complete the program then what happens? So the answer is actually job assist plus. Uh so there are four layers of career support that we provide. So first is technical capability enhancement. So assessments to sharpen your core skills. Interview prep with mock assessments tailored to your role. Uh then we have AI powered prof profile optimization. So simply learn helps you use AI to make your resume and LinkedIn profile stand out to recruiters. And we also have handpicked job opportunities to match your specific skills and goals. So this is not just a bonus feature. it is built into the program because getting you certified is just one aspect of what we provide at Simply. Uh and then coming to the most important part, I I received a few questions about the program fee as well. So for Indian learners, the course fee is $54,000 and you have the option to pay this in installments as low as $4,500 rupees per month. uh uh for learners joining us from the US it is $1,449 and you have the option to pay it in installments as well as low as $128 or dollars per month. Uh we also have attached a few testimonials from our uh course page. You can definitely go there and check it out as well. Um I will be quickly launching a poll for all of you. If you're interested uh to join the program or if you'd like to know more, you can definitely uh click on a yes and our team will reach out to you about the same. Uh so think about it uh and a few of the testimonials that we have included are from Alex and from Nina. So Alex is a business analyst in Chicago and he said that the program gave him a deeper perspective on modern business analysis and the evolving role of technology in decision making. Uh so that is what the experience was for him and for Nina. So she's a senior manager in Bangalore and she said that simply learns curriculum and flexibility help her balance work and studies as well and the certification helped her secure a senior manager role at access bank with a 40% salary hike. So these are real experiences uh from real um elite members. So we have a lot of testimonials that are covered as a part of this program. So please uh go ahead and give us your responses. We also come across a lot of people who are interested but they're not really sure if they want to join the program if this is the right time or some of them just require the right kind of guidance. So if you're wondering about it, if you're really not sure again these are uh course advices so you can probably talk to them uh and let get clarity on that. So uh please go ahead and give us your responses. I will also try to take a few questions. If you have um any please share it in the uh Q&A and I've seen that Solomon has answered a lot of questions. Uh a few of the other questions that are remaining on from our side is about the certificates. A few of you have mentioned that you've not received it. Some are facing issues like I mentioned please write to us at [email protected] and if you've not received the certificate after writing to us uh we will check the check with the team and get this resolved as soon as possible. So with that I'm going to uh end this poll right here. Uh if you are still interested and you were not able to see the poll or respond to the poll what you can do is uh you can go to the course page visit the course page and uh you will be able to talk there's an option to talk to the advisor. So you can do that and you will be able to uh get more details on this. And finally coming to the certificate uh pool. So what do you get along with this poll? uh the certificate the slide deck. The slide deck has a lot more content uh that we have not been able to cover because of the time limit that we had for the webinar. So all you'll have to do is fill your full name and your email ids are anyway recorded. So you will be receiving everything that I mentioned to your email id. But please keep in mind that you will receive this within 48 hours. So try not to u you know panic before that. Uh and if you're not able to see this poll, if you're not able to share your responses, again, please write to us at [email protected]. I've noted down a few names of the people who have shared their questions about certificates, some who have not received it yet. So, we will surely get back to all of you uh with your certificates, right? Um and yes, I think as you all are responding, I would also like to take this time to thank Solomon for delivering such a wonderful session. I understand that the time was limited and but it was great uh seeing you give the project demo. I'm sure everyone has clarity of what they can expect uh by the end of this course if they decide to join for the power AI powered business analyst course for uh with us. So uh Solomon before we wrap up I just wanted to quickly ask you if you have any qu uh you know um anything you would like to share with the audience. All right thank you so much for having me. It's been my pleasure being here. Yeah, I've also had a wonderful time. I just want to say to all of the learners that AI has come to stay. Yeah, I know there's a whole lot of talks around AI replacing jobs and so on and so forth. Yeah, but you need to get to break the barrier, get to start learning about AI, get to start using AI, get even to start learning how to build your own AI powered tools. Yeah, and that's where we come in at Simply Learn. Yeah, we able to help you get started in your journey. Yeah. So, I'll be looking forward to seeing several of you in forthcoming classes. Thank you. Thank you so much, Solomon. Uh that means a lot and I'm sure everyone benefited greatly from today's webinar. Uh so with that, I'm going to end this poll as well and we have come to the end of this webinar. It was wonderful having all of you as well. It was great hearing from you. So many comments, so many questions. So you guys were an engaged audience. If you have any further questions, if you have anything you want to share with us, any feedback, please write to us at [email protected]. And with that, I'm going to wrap this webinar up. I hope to see you all in the upcoming webinar. So once again,