Bad results in forecasting using CNN + LSTM

Hi, I'm a mathematician with a keen interest in deep learning (DL) and its various forms. I've been working on a personal project involving forecasting. Initially, I used ARIMA and achieved relatively good results. However, driven by my enthusiasm for DL, I attempted to enhance the predictions by implementing a combination of CNN and LSTM. Unfortunately, I'm encountering an issue where the model is outputting the mean value of the training input instead of meaningful predictions. This is reminiscent of …

arima cnn combination deep learning deeplearning forecasting forms good however lstm predictions project results