Grant: Engage Plus Grants Program

The focus of this work is the prediction of future behaviour based on past history for financial transactional*data. Much of the transactional data arises out of daily banking activities such as deposits, withdrawals,*interest, loan, mortgages and various financial investments. Increasingly, financial institutions are seeking to*provide enhanced consumer value as well as create competitive advantage by using transaction history as a*foundation for predicting future client behaviour in terms of revenue, cost and margin. This work is an*extension of the results from an Engage Grant with DecisionWorks, where the aim was to develop a*member-profile driven model for predicting future member behaviour, such as revenue, cost and margin. As a*result of the original Engage Grant tremendous progress was achieved toward this goal through introducing an*effective recurrent neural network (RNN) architecture, dataset feature selection, and a two-phase design of the*whole model using the defined RNN. The focus of this Engage Plus Grant is to evaluate the effectiveness of*this model in real life by performing a pilot experiment for the respective credit union that provided the data;*compare our RNN-based model with other existing models currently used by the industry; and integrate the*proposed solution into the DW's systems. The outcome of this grant will improve the Manitoban and Canadian*economy by giving Canadian financial institutions an edge over their competition in other jurisdictions and*create new job opportunities within the financial services sector for data science and business intelligence on*predicting member/consumer behaviour.