Abstract
Our topic focuses on constructing a stock forecasting system. This system can provide the user with the basic information of stock, stock price prediction, K-line chart, and the individual stock news. In the prediction aspect, our system uses four machine learning models, including Random Forest, XGBoost, LightGBM, and LSTM. To train our machine models, we use 250 different technical indicators as the variables and inputs. In order to make our model become better and more precise, we also used Shap and Skater the observe the reasoning process of the machine learning and improved our models by analyzing observation and changing parameters. We evaluate our models by examing the precision and recall rate and found that LightGBM has the best precision and recall rate. In addition, in the field of data processing, this study uses the network crawler to extract the share price of listed cabinet companies and combined with MySql access to individual stock data. Finally, we created a Telegram Bot to present our results to users. Users of our bot can enter several commands to our bot and our bot will read the information and return the result to users.
System Flow Chart
This chart demonstrates the basic flow of our system.
Applicatoin Flow Chart
This chart demonstrates the basic flow of our application for Telegram bot and output.
Telegram Bot Demo 🤖
For more details about the code and the structure of this research project, please refer to my GitHub Repository
This is a project supervised by Kao, Ming-Sung from Fu Jen Catholic University. I’m very grateful to him for his enthusiastic and responsible supervision on the project.