Abstract. In recent years, one of the most popular cryptocurrency assets that has attracted the interest of many investors is Bitcoin. Due to its increasing popularity and high price volatility, forecasting Bitcoin price has become a central concern in financial market analysis. We use machine learning techniques and SHAP (SHapley Additive ExPlanation) interaction values to improve our forecasting accuracy. The historical dataset used for this study includes various variables such as opening price, closing price, trading volume, and other factors. In this study, four different machine learning models are used to forecast the price of Bitcoin. The machine learning models used are XGBoost, CatBoost, Random Forest, and LightGBM. After the models are trained and tested for performance by evaluating metrics and visually analyzing historical data, the four models generate different price forecasting. To understand the contribution of each feature in the forecasting of Bitcoin, the SHAP (SHapley Additive ExPlanation) interaction value is employed. The SHAP value provides valuable insights into the extent of influence each attribute or variable has on the forecasting. The results show that utilizing machine learning methods and SHAP interactions can provide an overview of the importance of the relationship between existing features and how much influence each attribute has on the movement of cryptocurrencies, especially Bitcoin. Therefore, this can provide a more transparent picture and insight for market participants to make better investment decisions.