Graph-powered CRM
This is a modal window.
The media could not be loaded, either because the server or network failed or because the format is not supported.
Formal Metadata
| Title |
| |
| Title of Series | ||
| Number of Parts | 50 | |
| Author | ||
| License | CC Attribution 3.0 Germany: You are free to use, adapt and copy, distribute and transmit the work or content in adapted or unchanged form for any legal purpose as long as the work is attributed to the author in the manner specified by the author or licensor. | |
| Identifiers | 10.5446/55185 (DOI) | |
| Publisher | ||
| Release Date | ||
| Language | ||
| Production Year | 2019 |
Content Metadata
| Subject Area | ||
| Genre | ||
| Abstract |
|
00:00
Plane (geometry)Kolmogorov complexitySystem programmingAbstractionGraph (mathematics)Convex hullMusical ensembleCustomer relationship managementAnalytic setPhysical systemGraph (mathematics)Product (business)Graph theorySoftwareDatabaseMathematicsMoment (mathematics)Graph (mathematics)Food energyJSONXMLComputer animationMeeting/Interview
01:48
Graph (mathematics)Service (economics)BuildingDirected graphSystem programmingScale (map)Plane (geometry)Food energyTelecommunicationLocal GroupInstallable File SystemContinuous integrationConnectivity (graph theory)SoftwareStructured programmingArchitectureProcess modelingLevel (video gaming)Elasticity (physics)Device driverBipartite graphScalabilityCollaborationismTheoryDatabaseOpen setComputer networkOpen sourceMereologyCodeVideo gameLie groupBit rateMaizeWorld Wide Web ConsortiumProcess (computing)Link (knot theory)Variety (linguistics)Data storage deviceEnterprise architectureShift operatorMathematical analysisParticle systemPhysicsOpen sourceTerm (mathematics)Complex analysisPhysical systemGraph theoryBuildingComputer configurationInformationSet (mathematics)Relational databaseComplex systemProduct (business)Video gameSoftwareData modelData loggerMereologyPoint (geometry)Enterprise architectureLevel (video gaming)Computer architectureData analysisTurtle graphicsData streamCollaborationismExpert systemDatabaseMilitary baseGraph (mathematics)Symmetry (physics)Quantum stateSystem administratorUniverse (mathematics)Strategy gameSelf-organizationView (database)AreaSequelService (economics)Projective planePasswordFrequencyTheoryOffice suiteField (computer science)Link (knot theory)Data structureCustomer relationship managementStreaming mediaTwitterGraph (mathematics)BitPresentation of a groupDecision theoryVisualization (computer graphics)Standard deviationMathematical analysisConfluence (abstract rewriting)Touch typingOpen setComplex numberBus (computing)Arithmetic progressionStress (mechanics)Theory of relativityHydraulic jumpBeat (acoustics)Computer animation
10:53
Computer networkMathematical analysisTheoryParticle systemPhysicsPlane (geometry)Graph (mathematics)Vertex (graph theory)System programmingElement (mathematics)Data modelComplex analysisMathematicsRepresentation (politics)Database transactionAlgorithmPartition (number theory)AerodynamicsTelecommunicationCluster samplingReal numberPersonal digital assistantTerm (mathematics)Reference dataCustomer relationship managementIdentity managementOperations researchGraph theoryOptical disc driveFAQBoolean algebraData managementComplex systemProduct (business)Graph (mathematics)WeightSoftwareCentralizer and normalizerTelecommunicationInformationWebsitePhysical systemGraph theoryAlgorithmTheoryPoint (geometry)Food energyCASE <Informatik>Demo (music)Classical physicsStatisticsRepresentation (politics)Multiplication signReal-time operating systemBridging (networking)MereologyDifferent (Kate Ryan album)Type theoryGraph (mathematics)FacebookCustomer relationship managementTwitterStudent's t-testInteractive televisionDigital photographyCategory of beingElement (mathematics)Endliche ModelltheorieMathematicsTraverse (surveying)Online helpPurchasingNeuroinformatikMathematical analysisDynamical systemEvent horizonComplex analysisDivisorComputer animation
19:59
Mathematical analysisBoolean algebraPlane (geometry)Cone penetration testGraph (mathematics)Focus (optics)Computer networkProduct (business)Customer relationship managementView (database)Maxima and minimaData modelDemo (music)Similarity (geometry)Artificial neural networkEndliche ModelltheorieComplex systemPhysical systemPlastikkarteGraph theoryProduct (business)Design by contractData modelDatabaseInformationConnectivity (graph theory)TheoryReal numberBitRelational databaseDemosceneSingle-precision floating-point formatPoint (geometry)Graph (mathematics)FreewareStudent's t-testCategory of beingGraph (mathematics)Network topologyBoss CorporationLocal ringMathematical analysisProteinCustomer relationship managementCore dumpSlide ruleQuery languageDivisorFamilyData structureoutputDemo (music)Database transactionInstance (computer science)Directed graphLatent heatOperator (mathematics)File formatDialectWordProgram slicingNeuroinformatikSimilarity (geometry)SequelEvoluteHypermediaGame controllerMoment (mathematics)Uniform resource locatorMathematicsSystem administratorPlanningClient (computing)Ring (mathematics)Video game consoleComputer animation
29:05
Client (computing)Execution unitBraidString (computer science)Plane (geometry)Graph (mathematics)Landau theoryMIDIConvex hullUniform resource nameInterior (topology)Graph (mathematics)CodeScripting languageSample (statistics)Graph (mathematics)File formatForm (programming)Link (knot theory)Software developerFront and back endsArtificial neural networkGraph theoryComplex analysisAlgorithmFormal languageProduct (business)Physical systemFamilyDesign by contractSimilarity (geometry)Mathematical analysisGraph (mathematics)InformationWeb browserQuery languageCategory of beingEntire functionDifferent (Kate Ryan album)DistanceSet (mathematics)Row (database)CASE <Informatik>BitComputer fontMultiplication signTerm (mathematics)Point (geometry)1 (number)SoftwareData structureShared memoryArithmetic progressionInsertion lossComputer animation
36:25
Plane (geometry)Graph (mathematics)StapeldateiIntelGamma functionProduct (business)DistanceSet (mathematics)1 (number)Graph (mathematics)Mathematical analysisDifferent (Kate Ryan album)FamilyRepresentation (politics)Graph (mathematics)Uniform resource locatorInformationSimilarity (geometry)SpacetimeDialectPoint (geometry)Bit2 (number)Computer animation
39:52
Graph (mathematics)Mathematical analysisMultiplication signAlgorithmGraph (mathematics)File formatInformationVirtual machineStrategy gameString (computer science)Term (mathematics)Set (mathematics)Physical systemDatabaseSimilarity (geometry)Graph theoryCompilation albumKey (cryptography)Different (Kate Ryan album)Complex analysisSoftwareSemiconductor memoryDirected graphLibrary (computing)Client (computing)Query languageCASE <Informatik>Data structureNoise (electronics)Data managementInformation systemsProduct (business)Statement (computer science)Latent heatMachine learningLecture/ConferenceMeeting/Interview
45:03
Expert systemKolmogorov complexityPoint cloudInformation technology consultingAbstractionSystem programmingTurtle graphicsVideoconferencingMusical ensembleJSONXMLComputer animation
Transcript: English(auto-generated)
00:00
Thank you very much. Thank you for coming. Yes in fact today I'd like to focus on
00:21
adding some boost to our CRM systems thanks to graph analytics. We started in 2014 in fact by partnering with a product company based in Sweden called
00:40
Neo4j and since that very moment we started putting all our energies in spreading the knowledge about graphs not just about the problem because what I'd like to try to demonstrate today is that when we talk about something like
01:04
Neo4j which is again a graph database we are more talking about network science in general so there's a lots of interest around that and we will try to explain why we think that the world of network science is exploding today
01:23
more than ever and all the mathematics and the algos available since the very moment when Euler invented the graph theory are available
01:40
for us in order to analyze and try to discover new insights from our data so we try to understand how every different approach about having a different approach to data analysis based on the graph theory that is
02:00
based in Venice it was founded in 2004 15 years ago and now we are opening offices in many other areas especially in Pescada where we are building a competence center around graphs and in Roma and Milan because we have many customers there despite our idea and approaches to deliver
02:22
service worldwide in fact we are having many projects around the world as well and after 10 years we started focusing on data driven system, on building and helping companies in building data-driven systems because we
02:42
understood that all the complexity or most of the complexity and value around systems is in fact around data we might have heard that the concept of and the password of data is the new oil you know because they are becoming even and more and more important especially in this period
03:07
now if we are able to analyze and predict by looking at our data how our business should evolve should change maybe we could speed up and improve or for sure we could speed up and improve the business itself we are
03:24
working in many areas from the government to all the sectors this is just to demonstrate that the graph theory can be applied in very different scenarios now from banking to telco to retail and again the government as I
03:43
mentioned we have many skills in particular we are focused on all the no sequel technologies Neo4j is the graph database but we have other kinds of systems where we specialize and we are certified we are de facto IT
04:03
architects and we have a long-term experience in term of in term of building business critical systems when it comes to working with data validating high goals rather than rather than data analysis we think that it's
04:22
important to collaborate with the universities and we have we have many collaboration opened actually not only with Italian universities but also with the Technical University of Denmark where where we have lots of network science expert and researchers that as validating and defining new strategies
04:47
in applying the graph theory to to our business we are part of the Italian open source network this is especially why we are here because the
05:04
organizers red turtle are part of the Italian business network for open source we believe in open source of course our products are always open source despite there could be a license in order to cover the 24-hour
05:22
support we have a retard offer blown oh we have a second for elastic we have a symmetry for life ray and and many other companies belong to to this network and we all bring the concept that the open source is important also for for private and the public administration not just for
05:46
fun or not and and we aim to serve and provide the services at our enterprise level even if we are talking about open source we have
06:01
other partners in particular we are partnering with many players around around data couch bases are document database and for data graph database link is a graph data visualization tool and confident is the de facto standard or best known product when it comes to data architecture building
06:26
data data architectures and data streamings between between systems this is our part with with with Neo4j in particular we started and we are the first company in Italy covering covering and and working with this product but
06:46
let's jump a bit again bit into them into the presentation itself so first of all going data-driven brings real benefits this is very important because it's not just for special it's not just something that Larus is saying but it's something that was clearly stated by the
07:04
European Commission I will share the presentation and on that link you can see and read with me that companies that found the decision processes or knowledge about their data increase their productivity by five or six percent fact to six percent so it's and you can see the progression in in
07:25
the very very different in many fields and all of production so if you want want to become that the driven if you want to go real data driven we need to understand the data that we are managing today and it's interesting and
07:46
important to say that data have changed a lot compared to the past modern data are completely different to what we were used to managing with the
08:01
traditional relational databases first of all as you can see in this in this chart structured data are growing this is true but there the explosion is more about unstructured and semi-structured data in fact we today
08:21
we are forced and called to manage the text log files streams tweet and all the social networks now so it's important it's important to understand that today and even more in the future the kind of data that we will have to
08:40
manage are more about semi-structured and unstructured data rather than structured data this is why the no sequel movement has emerged and this is also why the traditional relational databases are are not the de facto anymore in fact relational database are more for governing and hosting and
09:06
managing structured data because you can you can you can you may know that you in a relational database you define your your structure you define your data model which is static okay and then you put the data inside it so you
09:22
are forced to predefine the structure of your data but when you don't have a structure because you're managing unstructured and semi-structured data then probably the relational databases are not the best option and not the best option because we need more flexibility the data are different you
09:42
can have articles you can have customers with very different information and not static as we are used to knowing and when it comes to working with data driven systems when it comes to managing this modern data
10:01
there's an another important aspect which is more about the network itself inside your data all the information that you are hosting are de facto connected they build the all the complex systems all the complex systems
10:21
are de facto based on on on a network and what we use to model complex systems are from a mathematical point of view are graphs so a set of notes and relationships between between them network theory as can be applied to
10:45
different and many disciplines from financial to telco to biological so to biology and if maybe we may know better about the social network now
11:04
Facebook and Twitter are de facto social network which are graphs of information people following other people people posting posting likes and putting likes to post etc. de facto social networks can be
11:22
analyzed via the network theory for sure no but you can also imagine that when it comes to a telco company the network the energy electrical network is of course a network or also and also for telco
11:41
when you have to to call from one point to another then you are moving the information and bringing the communication along a network no along a path in a network in particular so as I was saying complex systems are composed by many not identical elements and connected by different type
12:02
of interaction that we represents as graph it all comes from the network theory you might know about the seven bridges of Konigsberg where it all started with Euler the problem was about modeling a problem or solving a problem which was about walking all around the city and traversing just one
12:24
time all the seven bridges and starting from that Euler invented the network theory so you have the representation on the on the on the left on the right sorry so graph algorithm gives you a way to understand model and predict the
12:45
behavior of your connected data we are more interesting interested about understanding the system itself how the system works which are the dynamics inside these systems rather than doing classical statistical
13:04
approaches and and and algorithms because we want to predict and understand how the system is based it works how the information flows around around the network and we have many algorithm coming from the
13:21
network theory from the pathfinding where you can evaluate the quality of a path you can identify if a path exists between two nodes in your network which is very typical for example with Google Maps now when you have to go from one point to another you just compute the shortest path or probably the shortest weighted path which is a pathfinding algorithm from the network
13:46
theory and then you have other other algorithms which are important when it to analyzing your data the centrality so the role the importance that some nodes as inside the network rather than the communities understanding how
14:01
the information how people our entities are aggregated inside communities may help you to understand better the behaviors behind that and probably understand better how to react to that situation no we will see we will try to see that thing in the demo later when it comes to graph theory to
14:23
applying and using graphs we of course have some typical use cases from real-time recommendation because you have to you have to when you have to propose something for example into an e-commerce website you have to
14:41
understand who is looking for what and you have to understand his behavior inside the system what he both in the past what is looking for which kind of product which which kind of from which company he buys etc no in
15:10
order and you build a network to understand that and you can analyze the network in order to propose better recommendation to to the user you have
15:21
a master data management you can use a graph analysis in order to detect frauds the most common and famous event about fraud detection is the
15:41
Panama paper you you you you might have heard about that on there on the news and newspapers it came from a leaker they stole some some documents and tons of documents actually something like 12 terra data of data
16:00
were leaked and they in order to understand how people were able to put their money into those physical paradises they reconstruct the networks and how the money flowed along a network now because you you you
16:23
might imagine that that when you have to move your money from here to there then you have to traverse lots of and ask for help from lots of intermediate people other banks that can help you moving the data up until there
16:44
or no and they reconstruct that on a network and you can go and see and go and see and analyze the network in the website itself of the Panama papers then we have lots of other use cases which I just want to skip and deep dive to our our demo so which is the value to us in analyzing the data that
17:10
you have inside your CRM with a graph approach well of course of course we think first of all as I explained previously your your your CRM data are
17:28
a complex system you have lots of entities which are not identical from customer to purchases from products and so and you want you want to understand how could you perform better of course not the aim is to
17:44
analyze the data and try to catch and bring some values from the algos and all the computer mathematics in order to in order to try to understand better the data that you have inside your CRM because the more you understand
18:02
the data and the more you will be able to sell and which is which is the factor our target not in general we want to try to we want to try to understand the data and try to be more effective when it comes to offering some
18:23
products to our customer with the somehow being somehow comforted and supported by the data and probably probably we will we will we will be more effective in proposing some products rather than other no so first of all
18:45
the market which is a network again was was in the past analyzed by typical and traditional segmentation approaches for example we may segment our market by
19:08
by agent know some some some some boolears or some some particular values know in this example now we said we have segmented our market
19:21
between students between teenagers workers etc but we know that this is not working anymore because you you you you know that you know that of course between the category of students you may have very and many different
19:43
interests know some students some something is may have may have the idea of starting their own startup so so it's true we they have the same
20:02
age but someone want to open his own startup someone has other plans not so we need to understand better the behavior of people because because teenagers is a two-wide category no in order to to understand and propose
20:21
product when it comes for example to to a banker we might think that in order to have more teenagers opening opening a bank account we should provide provide a product that as for example free of charge for your mobile phones
20:45
no you don't pay anything in order to recharge to recharge your phone because we think that all the young guys and students are just interested in chatting with with with a phone which is how true but that's not all that's
21:04
not all for sure no because because among this category again we have to deep dive understand and analyze better the information that we have that we have inside the system because the data are there it's just a matter of analyzing them correctly no and and understand better all the information so
21:23
coarse grain segmentation by Asia for example is not a competitive advantage anymore for sure we need to focus on the customer we need to understand and better the customer we need to customize our proposals we need to customize our product we need to to go to a more personalized
21:46
proposition because because every people is different and we have to understand better what we we have to propose to them so your market is an customer products offers are linked for sure together no there's a person living
22:07
in some place moving some money to another person that person is paid with a credit card when he signed the contract all the information are somehow connected and this is important the graph databases and graph theory focus
22:25
more on the connections between information rather in the network structure rather than the single data point itself we are more interested about understanding how the ecosystem how the complex system works in itself no so let's think a bit about how we could model our CRM
22:47
information because if we are able to transform the information and put the information inside a CRM into a graph then we for sure will be able to immediately understand better the relationship between the end the
23:06
entities and and given that they are they become much more explicit then they are also easier to understand they are also easier to analyze etc we
23:23
can we could find unexpected interracial between between the data for example because we we can discover that the information that in theory should be disconnected are instead the connected for example in
23:40
your in your mind you have a sales guys covering some particular regions no so so rings is covering Venice and someone is covering Ferrara and so on and by analyzing the data we might discover that Lorenzo is instead selling also in the closer provinces because he is more effective or for
24:06
some reason no is selling more so we might discover an expected relationship between entities and then of course by applying the graph theory we could provide new insights between the data sorry for the Italian words
24:22
this is a real data model coming from one of our customer where we modeled we modeled the customer relationship system their customer relationship systems so it's pretty intuitive I think now everybody is able to read it up
24:41
apart from the Italian if someone is coming from not from Italy so you have you have a client a customer which is buying a product the red bull no which is belongs to a produce family the last bull on the other side and of
25:03
course you have an agent selling a product the agent as a supervisor okay which is which is a tree of relationship between agents up until the the boss no and agents works on for a specific agency which is based in a
25:23
locality in a certain province okay this is how the information de facto are connected in your system no this is how you could represent it is simplified of course your CRM probably have much more information so I know
25:41
but in general the core is that no you have again customer buying products which are sold by some audience belonging to some agencies in some places okay for sure you are missing the location and the province where the customer leaves no you could evolve the system and add even
26:04
more connection between information as I was as in the example I was doing one moment ago between Venice and Ferrara when it comes to selling by by locality province or region the good news is for no sequel databases and
26:24
graph databases as well is that is that evolving a model is much easier rather than evolving our relational database so if you want to add the relationship because we think it is interesting then it's just a matter of adding it without any change without any ticket open to our database
26:47
administration order to have this relationship not coming in in in the graph because they no sequel databases that are covering again structure and structure data better are meant to be more flexible knowing exactly in order
27:04
to cover this flexibility and and support the evolution of your of your system of your model of your analysis now so you can easily add and evolve the graph that is simplified here okay and then you could use it to
27:21
recommend the product to compute similarity between customer I will deep dive a bit more in the demo in particular in the second one and then you can do many other many other analysis let's have a look at them at
27:40
the demo so this is the graph console for Neo4j if we call DB dot schema that we could we could hopefully see the model that I just highlighted
28:02
putting the in the in the in the slides one second ago no this is the how the database is structured those are the entities present in the graph so you could create some nodes as agents you could create some nodes as
28:22
customers and so on and the relationship between between the instances of those entities represents real operation real transaction between those entities not for so for example we will have an instance of customer which could be Lorenzo buying the red the red arc buying specific
28:43
specific product belonging to a specific family okay this is exactly the structure that I was trying to and I represented in the slide one one second ago okay no let's try to let's try to run our very first
29:04
analysis or query this is for recommend recommending something no let's try what what it happened don't be scared about about the graph now I will reorganize it the tool is just for developers so it's not meant as a front-end we have linked curious as I mentioned before between our partners
29:24
that are specialists in represented representing data in this format in the form of a graph but in a business way no end user way rather than developer way as the Neo4j browser which is just meant to do some queries is meant
29:46
just for let's say data scientists that want to have fun with with and work not just fun and work with with our graph no here we have here we have two similar customers similar because as we can see they are buying some common
30:07
products so the red one are exactly the contracts so are different by definition now this is my contract for or my insurance policy or something like that
30:22
they are connected to the category no in order to understand that the two contracts are referring exactly to the same information in our system no so those are the produce itself or the families and those are the contracts the
30:41
fact oh no so we can see they have one product in common they have a second product in common maybe even more no so they are pretty similar because okay here we are no here we are they have four products in common okay so we are de facto computing a similarity algorithm now we are trying to
31:06
understand don't don't worry about the language which is not so important I just wanted to highlight that the intersection between the product they both has to be more than 66% no they they have to be pretty similar now we
31:24
are interested in understanding the behavior of customers in order to be more effective in proposing some products they could buy and that we want to sell to them no and here the interesting things is that those two
31:41
customer the blue balls okay are pretty similar six more than 60% of the things that we were able to sell them are exactly the same no okay well they have in fact four products in common but as we can see we have some differences in fact we were able to sell to the first customer those
32:04
two products okay those ones this and this okay and we were able to sell on the other side those two other products okay given on the network stuff given this network structure then we understand that we could be
32:25
effective more effective we will have more probability to sell something more to those two customer just by proposing to this customer what according to you exactly those two products okay because if we were able
32:45
to sell four products those ones okay to this customer okay and we know that those those products interested also to this customer in fact he bought them
33:01
as well okay it's very probable okay we have 60% of coincidence no in term of similarity between those two customers okay so we understood that they are very similar and so they probably have the same needs the same interests no when it comes to to buying something okay so at this point
33:25
we can try and we probably will be more effective by proposing exactly those two products to this customer and on the opposite by proposing those two ones to this other customer okay because they are very similar so we
33:44
understand that they share lots of interest lots of information no the similarity is a concept of the graph theory is it's a set of graph algorithms okay then what we could do
34:01
we could run a more a more complex analysis which in fact it takes much more time because now we are exploring the entire graph okay in order to compute what I did in a particular use case now because there in in the
34:20
previous example I just fixed three products okay or at least three products and I asked to find people sharing these three products with similarities over 66% now what I'm doing maybe I can enlarge a bit the font of course not zoom in of course okay okay so from this analysis we
34:57
understand that for example we have more than 4,000 customers which are
35:04
exactly identical because they just bought one product the mad best brands etc okay so when the distance is exactly zero it means then that this set is exactly similar okay not just similar but exactly equal no two
35:26
people both two products four people both the second row so three products okay we have almost 1,000 people just buying the euro pension tax benefit blah
35:40
blah blah okay so we could we understand that there are some communities aggregated around the same interest okay because they are sharing they are sharing exactly the same products they both exactly the same zero the distance is in fact zero zero then we have 11 people with a certain
36:05
difference 0.25 let's try to deep dive in these 11 11 customers I run the same algorithm so you just explore the entire graph in order to compute the
36:22
similarities okay and then I filter out I filter out them let me zoom out a bit that's okay okay again don't worry about let's let's try to let's
36:42
try to understand the one second a bit more about okay if I just take the product families okay and try to create some space in my graph then I immediately start seeing that there are some communities of customers around
37:03
around the produce okay so we of course we can see that the distance is 0.25 so they are pretty similar of course the distance was not so great so this is again interesting because if they are
37:21
pretty similar again we could probably try to propose something something to this those customers okay now more or less we see this community of customers because they are more or less sharing the same the same product but we have this unconnected this unconnected or this second second
37:44
community okay which is more interested about about buying these three sets this one this one and this one is for those customers while those ones are buying one two three and four customer products set of
38:03
products okay so coming to this representation then again we could analyze and try to understand which are the differences between those communities of customers in order to understand if it may could make sense to propose for example to these customers to those customers those sets of products
38:25
we should understand why we should understand why they are so they could be interested on that just by analyzing better the differences between the two communities it could be a problem of age it could be a
38:40
matter of rather than a problem of region it could be the just the sales guy is not proposing those products so we should say to the sales guy hey try to try to sell the four products there to that community okay
39:03
because the two ones are pretty similar so you have you may have more probability okay so this is basically all of course you can do geospatial and something interrupt me if you have questions feel free to of course I
39:24
could do some geospatial analysis just by by analyzing the look at the location or the where the customer lives and so on which could be another interesting information in order to compute the similarity I could be
39:41
interested in understanding if regions are similar or not and why they do they differ and so on okay so this is basically all so any question or curiosity yeah yes please hi I want to just know curiosity how do you gather
40:14
all the data and then find sure sure sure sure then it's just a matter of
40:20
importing the data into the graph no we have all the technologies the connectors in order to import to take the data from the blown database and and bring them in into a graph for sure you have to take the information which are inside the system and copy them into this graph
40:41
structure and after that you can do the the analysis that you you are interested in I mean if the information are huge probably you won't be able to do that let's say in memory with some Python library or something like that of course you can do that as well now it's just a matter of querying information in memory or bringing them into a graph as
41:04
I was demonstrating here and then just running the graph algos which might be a library a Python library for example or the graph algo it then and and present and natively into into into net4j for example okay any other
41:25
question when you import the data from many source sometimes the same client appear in the place how do you manage to okay to join those clients in
41:47
the same in the graph in fact that in fact I mentioned the master data management as one of the main use cases for applying and using a graph database because you you know complex information systems and big customers
42:06
typically as duplication of different silos where the information are replicated and of course when you want to do some this kind of analysis you need to identify correctly that this customer in the system is
42:22
exactly this one in the other system because otherwise the network won't be represented the effectively you know and it's an artwork it's an artwork because you of course need to clean the data which is something that we do at import time into into the graph database we try to do some cleaning we
42:42
try to understand if the if the customer is in fact the same just finalizing the name or some a set of information but nothing can help you in terms of this technology no it's just a matter of cleaning the data which is
43:00
unfortunately for the last thing you just say right now about data cleaning do you use any machine learning classification it may be it may help you for sure absolutely absolutely any any any any strategy any approach
43:22
to clean the data may may come to rescue no for sure because because because yeah you can do some similarities between strings you can do some NLP for example but of course you're right you could do also some machine learning and some AI algorithm know in order to understand
43:45
if the information are exactly the same what we do into sorry what we do into into our work into our graph when we import the information is try again to compute the similarities this time between the information that you
44:03
have inside the nodes no maybe you could understand that basically those two nodes are similar in terms of the information they contains they contain so so so but it's it's better as you are proposing for sure I
44:20
mean we have some merging commands statement in in in there for J with the graphs where the merge statements means try to find a similar or similar or exactly this node and if you find it use it instead of duplicating it no it's a specific command but but in order to understand that the key the unique key
44:43
that identify this is a customer of this product of course you have to apply all the strategies and again machine learners may help you a lot for sure other question okay the time is over thanks Lorenzo thank you for