ShouldAI

• 

  Home

• 

  Dashboard

Inspiration

As a college student, I have always been fighting to decide between two choices, either treating myself an expensive dinner one night or buying a new set of dev stickers for my laptop because I can only afford only one of them. I believe not only other college students, but also many other young people out there have the same issue. Therefore, I decide to make an app for this.

What it does

ShouldAI is there to help whenever you have to decide where or what to spend on within a budget.

How I built it

How I train a decision tree classifier: I collected merchants data from CapitalOne enterprise endpoint, and create sample transactions from customers endpoints and clean, extract features from there. Finally, I came up with normalizing input into type (food, entertainment, personal_use, other_expense) and labels are (useful,bad_use) Interface: A simple Angular2 app that asks user for 2 questions: should i do this or that, then processes the text using Google Cloud API's natural language processing to parse entities and syntax (verbs,actions) and send it to the classifier. Classifier normalize the input and predict base on the current model.

Challenges I ran into

I wasn't able to find someone within same interests so I had to handle everything by myself. Because I had to make up sample transactions data, it's harder to observe the patterns and the model might not be very reliable

Accomplishments that I'm proud of

I was able to collect data, extract features and train a classifier. I'm still developing the interface but I was glad that I challenged myself within 24 hours

What I learned

Practice data mining knowledge

What's next for ShouldAI

Integration with Siri IOS

Built With

• python
• scikit-learn
• angular.js
• google-cloud