Forecasting news popularity on social media platforms is a challenging task that requires accurate prediction. The news popularity progression needs to be monitored at an interval of specified time steps. The research was carried out using experiments on a real-world dataset of news articles. This problem was modeled as Multivariate Time Series prediction. The Popularity Index is monitored from three social media sources, Google+, LinkedIn and Facebook at an interval of 2 h statistical and Deep Machine Learning Models were used to predict the News Popularity Index after 48 h based on the progression of News every 2 h. For Statistical Regression Linear Regression Model was used. For Deep Learning Long Short-Term Memory based LSTM and BiLSTM models were used. The models were compared using the metric of root-mean-square error (RMSE) for the accurate prediction of News Popularity. LSTM and Bi-LSTM outperform regression techniques in terms of prediction accuracy. Additionally, the performance of LSTM and Bi-LSTM were compared. Bi-LSTM slightly outperforms LSTM. Our findings suggest that deep learning models, specifically LSTM and Bi-LSTM, are well-suited for forecasting news popularity on social media platforms. Our research provides insights into the strengths and weaknesses of these models and offers practical implications for media outlets and social media platforms in their efforts to increase the reach and impact of news content.