Stock Price Predicition

Problem Statement

Owning stocks could add a lot of value to one’s investment portfolio that could help in building savings and increasing/maintaining wealth, but like any investment one should be aware of the associated risks given the high volatility of the stock market. There are several trading approaches that could be followed by the investor being mainly short term trading and long term investment. Short term trading is usually defined as holding a stock for no more than a year, could be even a few hours, and focusing on the profit potential associated with volatility. On the other hand, long term investment is defined as holding a stock for more than a year and they tend to exhibit much lower volatility, and are more reasonable when it comes to taking decisions about investments such as the company’s performance or future plans. Stock price prediction is widely acknowledged as a challenging undertaking due to the extraordinarily unpredictable nature of financial markets. However, to make money or predict the trend, many market players and scholars use statistical, econometric, and even neural network models to forecast stock prices. Accurate stock market models can give investors the tools they need to make better decisions. These models can aid traders in minimizing investment risk and identifying the most profitable stocks. One of the most important aspects of temporal data forecasting is selecting the most promising algorithm for modeling and predicting a specific phenomena.We will explore different models that are used to predict stock prices and use evaluation metrics to find the most effective model that predicts the uncertainty of stock prices.

Dataset

Old Stock prices joined with tweets sentiment score on date

Input/Output Example

State of the art

Comparing with four widely used models for forecasting Deep Convolutional Generative Adversarial Network outperforms others in the evaluation metrics making it the best model.

Orignial Model from Literature

Proposed Updates

Update #1: 1. Feature Engineering: we combined multiple features in this model, which better represent the problem.In addition, the data is normalized in order to have data in close range. (A repeated suggested change in the feedback )

Update #2: We added another technique/predictor that has shown to have a very high influence on stock market prices and it mainly focus on tracking tweets of influencers such as Elon Musk, that I’ve personally witnessed his huge impact in the most recent incident when he tweeted his dog “Floki” with a caption “Floki Santa”, and we’ve seen the cryptocurrency name “Santa Floki” sky-rocketing in the next few hours to almost 100x, making new multi-millionaires overnight.

details here