COVID-19 has led to Termination of Employment for 15.6% of Employees from companies in Indonesia. The uncertain situation of work has urged the citizens to make a plan to save their financial situation. Though most companies have kept their employees through Working From Home, it is necessary to have an additional income considering electricity bills keep on bloating, supplies of primary needs when everyone is in quarantine. Here’s where the idea kicks in: fast money, easy to convert into cash, non-productive assets. Yup, seems like selling them would be the right way to go.

However, not everyone has the experience to know about property or automotive business and it could lead to unworthy pricing. Too cheap, you’ll get huge loss and regrets. Too expensive, no one would be interested in it. So what could possibly be the solution? Look down, the solution is at your fingertips.

Introducing, PredictSee: the ultimate application to estimate your non-productive assets’ selling price. PredictSee allows users to describe briefly about their cars and houses while still getting accurate pricing. PredictSee gives you an accurate insight of what’s next for your assets.

What it does

PredictSee is a user-friendly and powerful price predicting app for non-productive assets such as houses and cars. It offers an easy sign in using google account and users will have to choose between cars or houses to fill out a specification or facilities questions. Allowing users to give a brief description of their assets using the most influential aspects of houses and cars’ selling price. The app will process the data provided by the users and predict their worthy selling price. Not planning to sell your assets today? not a problem. PredictSee will give the objective selling price recommendation based on the quality of your assets. we’ve got it all!

How we built it

Java Android Firebase Auth Pandas Scikit learn google colab google spreadsheet Google script Multi linear Regression Figma Stats Model API seaborn

Challenges we ran into

  1. Time management is the biggest challenge of all because creating 2 machine learning models are very complex and took a lot of time.

  2. We created linear regression model using scikit learn and statsmode.api. this modelling process took a lot of pre-processing such as data cleansing, feature engineering, and train/test model.

  3. Scikit learn model needs to be deployed to android. The trick is to use spreadsheet as the medium. android apps application -> google spreadsheet -> scikit learn (google collab) XD

Accomplishments that we're proud of

We managed to finish the app and it works really well! Succeeding in deploying price prediction with great accuracy is probably the best feeling ever! Opening up our user-friendly app and using it is such a rewarding experience.

What we learned

  1. Its a first hackathon for our 2 members, the whole process are news to them

  2. We managed to have a really good teamwork without having to meet each other.

  3. we learned a lot about the mechanism of an online hackathon, such a great experience!

  4. Using Spreadsheet to bridge android and google collab to deploy machine learning model.

  5. Learned about multiple linear regression pre-processing for multi output data such as creating complex visualization data from various feature towards price parameters, feature selection using Recursive Feature Elimination (RFE) to get the best feature and model parameter tuning to get the best result.

  6. It’s very important to make a model based on a user-friendly feature data but maintain to have a good performance so the data that’s provided by users will be impactful and convenient to fill.

What's next for PredictSee

  1. Adding more assets to predict, maybe gadgets?

  2. making a machine learning model deployment to android apps using a much more proper tools

  3. Providing a comparison of the estimated selling price with buying and market price.

  4. Connecting users with an online market place, so they can directly sell and advertise their assets.

Share this project: