Budget Buddy

Inspiration

Looking through major issues in the financial department, we realized that 35% of Americans only have a few hundred dollars in their bank accounts, while another 34% have no savings at all. Due to the magnitude of the situation, we decided to plan a hack to help these people formulate a savings strategy for their money. Our Chatbot is especially designed for people with unstable jobs and finances, as they are an underrepresented population in other budgeting apps. These apps assume that their customers have the privilege of stable jobs with steady incomes.

What it does

Our Chatbot is just like a friend helping you with your finances. It is connected with the Messenger API, is able to take an input of words chosen by the user and pick out intent. Using Machine Learning and Natural Language Processing, we were able to give our AI many test cases of financially related questions, so that it could give users relevant information.

How we built it

We started by accessing the Facebook page and API to connect to our conversational UI. This UI is built from Webhook. We also used Flask to build the database portion as a storage for our data. We trained the Machine Learning aspect of this program using Wit.ai, a Facebook owned program which allowed us to find intent.

Challenges we ran into

While we were trying to set up our project, we realized that a database would be necessary to store information for recall. However, when we tried to set up this database, it was set as a read-only file, rather than a read and write one. We attempted to Google the issue and look for credible sources to help us with the issue, but nothing we tried would work. We tried many ideas after we were unable to debug the code, such as creating a TXT file to store the data. However, this TXT file did not work for our project, and we realized that a database would be the only way. Scared we won't be able to finish in time, we tried a last resort. One of our team members was able to use Flask to bypass the bugs and create the database.

Accomplishments that we're proud of

We are proud of training our own machine, with the ability to distinguish between buying, selling, earning, and changing our budget. We also taught it to recognize the differences between food items, clothing items, transportation, and other miscellaneous expenditures.

What we learned

We learned how to learn Machine Learning and Natural Language Processing in conjunction with a conversational UI to create a savings plan for people in need.

What's next for Budget Buddy

The next steps for Budget Buddy is connecting to Capital One Bank accounts so we can keep more accurate records of spending and help create more accurate saving strategies. Also, we want to be able to expand our functions to include opening a new credit card, or helping with taxes to name a few.