Predicting Financial Savings Decisions Using Sigmoid Function and Information Gain Ratio

Abstract Planning for savings remains one of the most critical decisions for any user. The most important factor in this process is decision making. Most of the time, we can take decisions on how much money to save at a time based on our spending pattern. But automating this process is not easy, since it involves a number of parameters. Here, we attempt to incorporate intelligence into this decision making. The algorithm will attempt to predict the maximum amount to save based on the current account balance, clubbed with the entire database of available transactions on that account / user. Every transaction will be assigned an impact factor based on the time of occurrence, relative to the current date. Every month is divided into four quarters to track recurring expenses like EMIs. These impacts have to be taken by a machine learning algorithm to predict the maximum possible savings in that quarter. The impact factor will also depend upon the fraction of balance being spent on that quarter. If there is a goal set for savings, it will also be taken into consideration. If a considerable expense is predicted for that month, the savings will be kept low so that the account won’t go into overdraft. Recurring expenses are kept in check and accounted for to the maximum extent using information gain ratio from the transaction list.