Change Point Detection and Kernel Ridge Regression for Trend Analysis on Financial Data

Abstract: The investing market can be a cold ruthless place for the layman. In order to get the chance of making money in this business one must place countless hours on research, with many different parameters to handle in order to reach success. To reduce the risk, one must look to many different companies operating in multiple fields and industries. In other words, it can be a hard task to manage this feat. With modern technology, there is now lots of potential to handle this tedious analysis autonomously using machine learning and clever algorithms. With this approach, the amount of analyzes is only limited by the capacity of the computer. Resulting in a number far greater than if done by hand. This study aims at exploring the possibilities to modify and implement efficient algorithms in the field of finance. The study utilizes the power of kernel methods in order to algorithmically analyze the patterns found in financial data efficiently. By combining the powerful tools of change point detection and nonlinear regression the computer can classify the different trends and moods in the market. The study culminates to a tool for analyzing data from the stock market in a way that minimizes the influence from short spikes and drops, and instead is influenced by the underlying pattern. But also, an additional tool for predicting future movements in the price.