Stock Forecaster

Glimpse of Project

Inspiration

When we were brainstorming ideas for this project, our main focus was to develop something that would be problem solving. We discussed a lot until Goutham came out with this idea to create really efficient algorithm for stock prediction. It was also inspired by the passionate interest we all carry for stocks trading, and will now inspire maybe many more people for stock trading with this idea.

What it does

Stock Forecaster asks you for any stock name you wish to invest in and it gives you the graph of this stock for the selected period and gives its predicted value for the stock for the next 15 days. We even tried several tests for our algorithm to check its strength and it gave out a result that has 98% accuracy. You don't know anything about trading? We have a designated terminology that explains each term with detail and helps you get started. Still have doubts? Ask our chatbot any stocks related questions and it will help you with almost everything. It also gives out a summary of the original data and the prediction for trends related to the company. We also have a note keeping section where you can save your work and come back to later.

How we built it

Goutham gave us a really good start by planning out the project precisely. We started off by extracting the data from Kaggle in csv format. Anita helped in Data Cleaning for the csv file and got it ready to be processed. Sahib and Bhuvesh then tried to process the data, where Bhuvesh tried to design a particular formula to compare the price changes of each month and correlate it to find future trends, but Sahib took a different approach to build the prediction method by using Rigid Regression method which he learned in one of his classes, and before predicting the future values, we tried to train our project on our actual data set values. Once we got good accuracy on actual data set, we went to predict for the future with varying sample period with the help of AI, and finally got good results by keeping the prediction unto 15 days. Then once we were done with it, we switched to build a website to showcase the project. We started off in Streamlit by providing a basic interface, improving it gradually by providing the basic structure of our code in Python. We then using AI, implemented the chatbot and test and train method in real time. We also used web resources to design different chart styles for our graph. We implemented features like notepad, and dataset on our own.

Challenges we ran into

When we first extracted the data, we thought it would be better to do the data cleaning and processing in R, but we soon found out that we had data losses when we tried to implement R with HTML/CSS, leading us to continue with Python. We then faced problems in finding an efficient algorithm that would give out a good accuracy to our results. Thirdly, bringing the processed data into our website was really challenging and ultimately we had to use AI to take us through.

Accomplishments that we're proud of

We are proud that we made a plan for this and stuck by it, following it passionately. Even when one of the team member felt lost, others were present to lift their spirits up. Helping and Communicating with each other through the challenges in their tasks was a special quality about our group. We are also proud that we were able to design such an algorithm that once felt infeasible. We are even proud at our problem solving skills we displayed, facing each obstacle head-on. Lastly, even after the lack of sleep during the 24 hours, we are proud that we finished our project in the given frame successfully.

What we learned

We all learned a variety of skills throughout this hackathon. From attending the first Git Workshop to completing the final piece of code, it has been an amazing learning experience. We learnt skills like Data Cleaning, Data Processing, Web Development, Optimizing AI, and most importantly a practical experience to finding a solution to a real world problem.

What's next for Stock Forecasting

This felt like a heavy job, but we are convinced that this can grow into a potentially sustainable website. In the future, we would like to test even more algorithms and formulas on more datasets. We would try to broaden our website as the time didn't really allow us to implement some features for the website. For example, we could not implement the chatbot the way we wanted it. We would also like to have a news portal on our website which would cover the news that could potentially affect stock market.