Analysis and prediction information system development for brokerage company

In this thesis we build a prototype of an information system, which enables us the performance of various analyses and predictions on the data of brokerage firms. Within analyses and predictions we implement six key functionalities: portfolio return rates calculation, calculation of average return rates of financial instruments before transaction execution, portfolio volatility calculation, account lifetime value calculation, predictions about whether accounts are to be closed and clustering of accounts. For the purposes of the last two of the enumerated functionalities, we use machine learning methods: we use random forests algorithm for predictions and hierarchical clustering algorithm for clustering. We demonstrate the efficiency of both algorithms on real data that was provided to us by one of the domestic brokerage firms. We also create two databases to store the data on which all the functionalities are being performed on. We also implement the automatic data-transfer process, which transfers all of the data from brokerage firm’s internal database to our databases. In the end we build a user interface in the form of a web application, which uses charts and tables to display the results of analyses and predictions