Fake news is a pervasive issue in the digital age, exacerbated by the rise of social media platforms. Traditional methods of detecting fake news, relying heavily on manual feature engineering and classical machine learning algorithms, have proven inadequate in addressing the dynamic and complex nature of modern fake news dissemination. In this paper, we present the work performed to develop a model for predicting fake news classification. Two stages were involved in constructing the suggested model. The term frequency-inverse term frequency (TF-IDF) approach was used to describe the characteristics that were taken from the news material and pre-processed using n-gram in the first phase. An ensemble sequential deep learning network, LSTM-RNN model, or HLRNN model was used in the second phase to extract hidden features to categorize news genres with accuracy. The findings show that the hybrid LSTM-RNN approach (HLRNN) performs quite well when tested over 25 epochs and 64 batch sizes with a maximum accuracy of 99.38% and a minimal loss for both training and validation. Moreover, testing loss and validation loss obtained by the HLRNN model outperformed the traditional RNN and LSTM model.