Predictive Analytics for Stock Prices Using Machine Learning Techniques

This dissertation examines the use of advanced methods of machine learning and statistical models that can be used to handle stock price prediction and market trend analysis. The ARIMA and Prophet models will be applied, together with unsupervised clustering approaches like K-means, to derive insight into the short-term prediction of stock prices and long-term trends. The researcher used huge stock price data from around the world to train and test their models, keeping special care for industries and stocks exhibiting both linear and nonlinear behaviour. This paper also informs in detail how time series forecasting is subjected to many challenges along with seasonality and volatility, even due to some external factors like holidays and big financial events. Each one is ranked according to the rigorous evaluation by MAE, RMSE, and MAPE metrics. The results bring out the strengths of ARIMA in stable market conditions and Prophet’s relative strengths in handling complexity with strong seasonality. Besides time series forecasting, the current dissertation also applies a clustering analysis to group the stocks by their volatility and performance task giving investors a more profound insight into the risks and opportunities of the market. The results of this study have pointed out the fact that such a combination of statistical techniques with machine learning algorithms significantly improves the accuracy of stock price predictions and subsequently yields better decision-making by investors. The research has added to the ever-growing domain of financial forecasting in a way that has identified how practically viable machine learning models are in stock markets while simultaneously developing a skeleton for further research in predictive analytics.