Did you know that 73% of Americans rank personal finances as the number one stress in their lives? This issue is only getting worse, as 82% of Gen Z and 81% of Millennials say that finances cause them stress. To mitigate stress due to personal finances, customers need to be more informed about their spending habits. Prediction of future spending is the key to gaining full control over current spending.


What it does

Spendli predicts future spending by analyzing past spending habits. The user can log into their dashboard, where components provide an organized visualization of their finances. Using machine learning, Spendli provides the user with a prediction about future spending to help users understand their habits. Additionally, credit risk is analyzed and displayed for the user.

How we built it

  • The front-end was built with Angular.
  • Azure ML Studio was used to create boosted decision tree regression and k-means clustering machine learning models to train the dataset for predictions and risk-assessment.
  • Python with the Facebook Prophet Library was used to predict data, and Flask was used to send data to Angular.

Challenges we ran into

The dataset we used did not have transactions for each day of the month, and was therefore not ideal for this application. Additionally, due to time constraints, our model was trained quickly. As beginners with machine learning, our group had trouble figuring out the right models and approaches to solve our problem.

Accomplishments that we're proud of

  • We were able to implement predictions of user’s spending habits for the next fiscal year.
  • We were able to process data to provide risk assessment for the users.
  • We built a beautiful UI to display our results in a useful manner.

What we learned

  • Use of Azure Machine Learning Studio to easily create machine learning models.
  • Use of Facebook’s Prophet ML Library to predict future data.
  • Using Flask server to interface between back-end and front-end.
  • Difference between Supervised and Unsupervised Machine Learning.
  • Use of Excel Functions to clean the dataset.

What's next for Spendli

  • Extend the application for B2B use by creating options for enterprise users to analyze data for multiple customers.
  • Improve the Machine Learning model’s accuracy to better represent the habits of the users.
  • Add features and perks to make tracking spending more fun.
Share this project: