Neo4j Bloom: Investigating Patterns in Financial Transactions

Neo4j Bloom: Investigating Patterns in Financial Transactions



hello everyone my name is Rick Rick I'm Ruben from near 4j and Here I am again recording another demonstration of the new fancy bloom new J bloom visualization and discovery tool set and today I won't be demonstrating you this fancy new tools at using a peer database but I'll actually be using a little bit of more sophisticated and interesting data set at least for business users and that's a fraud detection data set this fraud detection data set has been generated for for us right so it's a fake data set but it's still quite interesting it allows us to look at specific types of patterns that we see a lot in the fraud domain so let's get going with that so I've got my frog my bloom demo going here but as you can tell it has an old perspective so I'll probably need to generate a new perspective here right and I'll use the auto generate capability to do that you have to generate this capability actually looks into the debate into the database and it finds the different labels and types of data that I've got in there and it assigns some default coloring to that I can change that coloring in the way I want and actually what I can also do is I change somebody a looking feel aspect spell go get into that a little bit later so now I can start interacting with it so if I for example go here and I say I can't hold her right so then I click this and I get a lot of account holders right more than five thousand nodes displayed on the screen and you'll notice that it does that quite quickly and quite easily I can zoom in the way I want right I can do whatever I want I can select and unselect by clicking over here lets unselect a little bit a little part of it and then go back here and Siri select a smaller part of it I can do that by drawing on that rectangle right a right-click or wrong click I mean and then select like that now it's selected part of the data set right and I can see that because if I do command shift D it will only leave that part of the raisers open command F right and it shows me those data elements in the center of the screen I can do all of that with the mouse as well rice command F fit to selection this Mesa branch of nodes or shift command T and that dismisses the other nodes so now it's um it's given me these these nodes right and I can actually start navigating right I can say okay why don't you give me a couple of these right for example this one that one and that one right so now there's three of them that are highlighted with a white little band at the age right and I guess a control e expand those show me what's inside right I can expand those again by doing another control E and another control E and then you know you can see that the data set gets richer richer richer ctrl shift D and it dismisses everything else again and that's how I can start looking at this in quite a bit more detail right if for example let's see here I've got a a little bit of a problem in one of the data elements here right so I'm zooming in here and all of a sudden I noticed that hey this guy over here has a wrong spelling rights on the v8 or that's clearly not the right name it's supposed to be shown the VI Torah right so then I can say control I or right-click and inspect right and it gives me more data about this particular node gives the birthdate first name full name last name and all of that stuff right and I can say okay I want to change this property right the full name right save that back and of course also change this phone because you know we want to keep it consistent right now this has also been updated in the data set all right so that's that's kinda kind of neat you can also edit the data obviously only if you have the right permissions on the data set right so that was part of this right but I can also have some other patterns here if I want to say first name Dirk right then you can tell right it's touring right is actually creating a more sophisticated graph pattern by associating Dirk to the first name property of the account holder nodes right it's finally one node that's Derrick stakes here right and so I can actually start looking at this guy and expand this is some more right you also notice that sometimes when you have a lot of nodes on the screen I'd like in in this case again right then I have all the account holder notes making kind it can get kind of complicated to see what's on the screen right and sometimes you actually want to customize the visualization a little bit right so I can go into the perspective and say that hey account holders are actually persons right so what I can do is I can associate a little I can do that I can do the same thing with a bank account and say okay this guy wants to have Bank I can write card you know I'd like to have a card I can I put that one a credit card that should be a different type of icon all right so I've got the Amex here let's see what else do I have here our social security number that's an interesting one usually right so I associate another I can with that as well and then of course the phone number right I want to have the last and not least I'd like to have a specific icon for that one as well right now it's cool about this is that if you look at it this way there's nothing much to be seen right but if you zoom in all right let's put this in the center here you can see that all of a sudden the I can start appearing right and you zoom in some more and then the captions start appearing all right so it does a really nice job but you know creating these visual cues for you to understand you know what's actually displayed on the screen I like that a lot they think it's very very powerful right what else what's also really powerful is of course the query capabilities that we find in the perspectives right so we've looked a little bit at the graph patterns already but the real power in my humble opinion is the custom search phrases so I'd like to show you how that's done and how we can create these custom search phrases really really easily now in a frame fraud data set I'm going to create a search phrase that is looking for five for all rings right so rings and to do that I've got a really complicated cipher query just a copy-paste here which is doing exactly that it looks for the four rings right now as a business user I don't want to be confronted with these for all the fraud rings queries right I just want to be looking at fraud rings nothing more I'll add one more here like a skimmer query right finding card skimmers find skimmers all right I'll save that and now if I go back to my perspective and say okay I want to look at the fly find the fraud rings right it's actually going to show me four nodes that are involved in some kind of fraudulent ring structures right doesn't really show here but if I expand these guys right then very quickly it becomes obvious that these guys are doing fishy things right not only are they sharing social security information they're also sharing phone numbers you know some of them may be sharing addresses right as you can tell over here so there's clearly something fishy going on here I don't do the same for the skimmers card skimmers right and it's actually showing me that really really easily as well right so it's got the skimmers right here let me select that fit that screen you see this card over here and it's been using different transactions and different people and it's it's immediately showing me that now this custom search query functionality is actually even more powerful than that because it allows you to parametrize these queries right what do I mean with that well it can ask you can ask the user for interesting input right it can you can say for example if you were looking for fraudulent potentially fraudulent transactions above a certain threshold then obviously it's quite easy to write a query for that I where you match a money transfer and its surrounding nodes where the amount of the money transfer is based on a user parameter right I can actually create a search phrase here that says money transfers above let's just say above and then say parameter right so this is the parameter that you find over here money transfers right the parameter is going to be offering up some suggestions suggestions are can can be done with these types of queries right based on the money transfer label and it's key right it's going to offer up some some some potential suggestions so what I need to do now is you know in typical business language money transfers above a certain level right so if I say money transfers above I don't know five hundred right it's going to give me a whole lot of notes still it's still quite a bit of information here difficult to to look at that so let me clear this one again and say money money transfers above I don't know twenty thousands right and then it's probably going to show me a little bit less information and something that is easier for me to investigate right so if I now look at the money transfers rights or the money transfer so that the yellow ones here all right let's select those dismiss the others and maybe select a couple of these guys and expand them right then all of a sudden we can start investigating what's really going on here that's really the power of these custom search phrases there's a lot more to it I think blue is doing a great job at explore exposing information in the graph structures to us in a very very user friendly and business oriented way and I look forward to seeing more of that in the future and I wish that you would all explore this as well you know there's a lot of things that you can do with it and maybe it's applicable to your domain as well hopefully so I'll look forward to hearing from you and I wish you a fantastic

4 thoughts on “Neo4j Bloom: Investigating Patterns in Financial Transactions

  1. This tool is just amazing! I always wanted a more powerful visualization tool for graph databases. And now staff members can write simple queries directly in english! Beautiful.

Leave a Reply

Your email address will not be published. Required fields are marked *