This project will build a model over a simulated data under a typical bureaucratic and political scenario. The simulated data will be stored in the Neo4J graphical database.The model will use the method of Deep learning (Machine learning) to train over the data and produce prediction algorithm. The model will systematically study the networking and activities of the member up to the 10th degree of relationship in the graph. For example, a politician's relative/friends get an abrupt increase in property value will add the probability to the politician be corrupted.
Simulation : This project is based on simulated data of people working in the current beuracratic and social structure in the country. Once a data structure of all major brances (e.g. government officers, politicians, teachers, professors, business persons, security officers) are created, a randomly generated purchase/transection/expenditure data will be created. This data will include message, bank transection, daily expenditure etc.
Implementation of Graphical Database (Neo4j) : The simulated data is rich in network and graph structure. Neo4J provides a best platform to operate over such graphical data. The simulated data can be directly visualized using ne04j API and numerious graph algorithm can be implemented to investigate over sample data. Smoe nodes(persons) are intentionally made corrupted in the simulated data by adding extra expenditure pattern.
Deep learning : Graphical data can be converted to vectorized data by implementing node2vect. A node will be represented by a vector and deep neural network will be created to train the model over labeled data. Once the model is ready, we can test the performace of the model by test data.