About Project

I started working on Thrift at Hack The North 2019. This project is about optimizing how people can save money through a web application that connects with your bank information through an API. Lets say you want to save X amount of money in Y amount of time. The web app analyzes your spending habits, income streams, and priority payments in order to give you a close estimate of whether it is possible or not to save that amount of money in the given time. If it is possible to save that money, the web application provides a monthly savings plan where it makes budget cuts on spending categories based on their priority.

At the core, this project is a constrained optimization problem that can be solved using Linear Programming. Priority payments such as tuition, medicine, bills, etc. all act as constraints that must be satisfied each month. Spending habits such as entertainment, take out, gifts, etc. are all variable expenditures that can be optimized. The optimization problem can be stated as following: Given a monthly income, what is the minimum amount of budget cuts that we can make to our spending habits such that we can save X amount of money in Y months and still afford priority payments.

How the Project was Built:

As mentioned, the project was a web application. The main components of the project are:

• Django for Backend and Financial Analysis.
• React for Frontend.
• Bootstrap as a Component Library.
• PuLP as a Linear Programming Library.
• TD's DaVinci API for bank customer data (no longer exists)

The main component that allowed this project to be possible was the availibility of TD's DaVinci API which included a database of 1 million + randomly generated customer data. Unfortunately, the DaVinci API is no longer active, so a major portion of the web application is no longer available.

Python was used to perform analysis to find reoccuring patterns in customer payments. Each customer had a history of payments available and patterns in spending were established based on monthly spending. This analysis identified consistent income streams, priority payments, spending habits, and some estimate of miscellaneous payments. The longer the customer history, the more accurate the identified patterns were. The analysis also identified all the categories the customer makes expenditures on such as shopping, dining, sports, etc.

Next was solving the constrained optimization problem. This was done using PuLP, a linear programming library available in Python. Each user had the ability to insert some weighting on their spending habits (eg how much do you care about entertainment, take out, etc. ). These weightings impacted the budget cuts in the sense that higher budget cuts are taken out of low weighted categories if possible. The Linear program then takes into account all the category spendings from the Financial analysis, the category weightings, and reoccuring payments and outputs an answer. It can output no possible if the income streams are not sufficient for the amount of money being saved. This can be adjusted by adjusting the savings period, savings amount, or changing the weightings on the spending categories. Otherwise, it will output a monthly savings plan identifying the amount of money that could be spent on each category, including miscellaneous payments. Of course, the output is an estimate so it gaurantees to get you close enough to that savings amount, but necessarily the exact amount.

Finally was developing a front-end for all the calculations that are occuring. This was done with Django + React. A REST API was developed in Django that calculated savings plans per user request on the client side. A dashboard was developed to represent the customer data and output the tables that contain the saving plans for customers. Some parts of the dashboard are represented in the images bellow.