Dublin High ScoolClass of 2024Dublin, CA
- "Can a machine learning model predict IBM stocks in the incoming years?" with mentor Leif (Oct. 3, 2022)
Anay's Symposium Presentation
Can a machine learning model predict IBM stocks in the incoming years?
Started May 5, 2022
Abstract or project description
Predicting future trends in stock prices and analyzing these trends is quite difficult in its implementation. There are many factors involved in the implementation that cannot be considered such as physical or psychological behavior. However, the goal of this project will be to have a fully trained machine learning model that can forecast future trends in the International Business Machine(IBM) stock in ideal conditions. This project would be written fully in python where it is not difficult to use and train a computer with data sets. The main objectives of this project would be data harvesting, data cleaning, and training of a machine learning model to get as accurate as possible to predicting future prices for Tata Global Beverage. In specific a technical analysis would be used to measure intrinsic values of the stock and a fundamental analysis will be used to predict trends in stock price as well as volume. The data set that I will use will be from the Nasdaq website which has the historical stock data for IBM from the last 10 years. Many techniques such as moving average, Prophet, Auto Arima, kNN, and LTSM, would be used to aid in the completion of this project. Simple techniques such as moving average will be used by taking the latest “close value” for each day in the data set and while considering the predicted values, the oldest value would be removed from the data set. Another technique that would be used is Prophet which is a time series forecasting library that tries to capture the seasonality in the past data. Auto Arima is another statistical forecasting method for time series forecasting which takes in past values to predict future values that will be tested in my project. The next method that would be used in my project would be the kNN or k-Nearest Neighbor method which takes the nearest data points and uses those to predict future cases. Lastly, I would be testing LTSM or Long Short Term Memory to store information that is considered important to the model and forget information that is considered is not. Overall, through using these techniques I want to be able to have a working model that can function to predict stock prices as best as possible in the incoming years.