Investment in sharia compliant stocks has experienced a significant rise in prominence in predominantly Muslim countries, including Indonesia. This is attributable to the fact that the sharia stock market follows Islamic principles, which Muslim investors must consider while making investment decisions. Sharia stock price forecasting is significant in this regard since it provides information for making investment decisions and limiting risk. While traditional techniques have been used to smooth short term fluctuations, machine learning models offer improved precision and flexibility. However, stock price data is frequently characterized by noise and prone to unpredictable fluctuations, obscuring key patterns. In this regard, wavelet transforms effectively filter noise, enhancing data quality and forecasting accuracy. Building on these insights, this study aims to evaluate the performance of various sharia stock price forecasting methods, both with and without wavelet denoising. This includes statistical approaches and machine learning approaches. Furthermore, we also employ a hybrid approach and propose a modified technique that combines CNN LSTM and CNN Bi-LSTM with wavelets. The combination of CNN and Bi-LSTM enables improved understanding of spatial and temporal data, such as the classification of time series data. The results show that wavelet integration outperforms the statistical, machine learning, and hybrid approaches.