To keep the things simple, it’s good to involve the people to manage the process. Process is managed by team based in Europe and India. There are two kinds of processes
- Rule generation process managed by data scientist
- Rule management process managed by rule engine expert
Rule generation process using Fraud Analyser: is managed by team of data scientists. One dedicated data scientist is assigned to each project. The data scientist is trained on company’s fraud management framework and has over 3 years of experience of building machine learning model. He is mentored and supported by senior data scientist and other senior management people. An ideal data scientist should have the following qualifications and experiences:
Quantitative skills: Data scientist should have background in statistics, machine learning and data mining. He should be able to analyse the data using quantitative techniques and find the business pattern as required. He should be well advanced to figure out the machine learning technique to apply under different situations. He should be analytical to understand the business problem and represent results in appropriate way. He should be able to select best techniques to solve the problem at hand.
Good Programmer: Data scientist should be good programmer. They should be able to manipulate the data as required as part of data pre-processing. He should be able to develop the model manually to benchmark the performance of fraud analyser. He should be able to perform additional activity such sensitivity analysis, model deployment and back testing. He should have solid programming skills to automate the repetitive task.
Communications skills: Data scientist should have excellent communication skills. He should be able to explain the problem to business users in laymen terms and same time able to explain the technical staff in way that they understand. He should be able to take inputs from both business users and technical staff and execute the tasks as required.
Visualization skills: A picture says things better than words. A data scientist should have good data visualization skills to understand the data and share the understanding with client as required. He should be able figure out fraud pattern by visualizing the data. He should know how to represent the analytical model in way that business user can understand.
Business acumen: A data scientist should solid knowledge of business to understand the problem, so do analytical work accordingly and spend time and energy in most efficient way. He should have good understanding of regulations and environment to deliver the projects accordingly. He should be able to apply new tools and techniques in new business setting according to client’s requirement and justify it.
Rule management process using rule engine: Once rules are generated using Fraud analyser by data scientist, they need to be implemented in rule engine or MS Excel depending upon requirement of the clients if they want instant results of fraud detection and prevention process. If it’s case of claim processing and there is lead time of 1 weeks, then it’s possible to deliver the results by using MS Excel and with support of data scientist. There is no additional resource required.
However, if it’s case of application fraud, immediate results are preferred. In case, there is no fraud then immediately application is processed. To get the instant response, rule generated by fraud analyser is implemented in rule engine. This rule engine is integrated with application engine. The diagram (source: internet) below gives the idea of the process with IBM’s rule engine- IBM ODM.
There are following components:
Rule Generation Layer: is managed by data scientist using Fraud analyser. He is business rule developer/business professional.
Rule Engine Server: one of the rule engines provided by IBM, FICO and Radhat can be used.
Java Server: The server is integrated with rule engine and web application. Rule hosted on rule engine are executed on Java server.
Web application: talks to Java server for each application decision using Java and XML.
Rule Engine Architect or expert: As mentioned before, rule engine side is managed by rule engine architect or expert. He is IT specialist with business knowledge of rule management and has solid Java skills. He is responsible for implementing rules in rule engine and integrating rule engine with Java server. Additionally, he is responsible for integration of web application with Java server.
Please note that rule engine management is additional service provided by our team at additional cost.