The prediction of stock movements has gained huge interest amongst the business world and academia. Due to globalization of advanced information technologies, the majority of peoples look forward in stock markets for earning huge returns. This paper presents stock market prediction using optimization driven deep model. Here, the review data of user and stock data are fed as an input to feature extraction. The extraction of features is done from the stock data and review data. From stock data, the technical indicators, like Absolute Price Oscillator (APO), Directional Movement Indicators (DMI), Simple Moving Average (SMA), WILLIAMS’s % R (WILLR), triangular moving average (TRIMA), and percent price oscillator (PPO) are extracted. From review data, the stemming and stop word removal methods are applied for extracting the irrelevant words. Then, the SentiWordNet is applied for extracting the review features. The obtained features are modeled in the feature vector and fed to deep convolutional long short-term memory (Deep-ConvLSTM) for stock prediction. Here, the training of Deep-ConvLSTM is performed by newly devised optimization technique, namely Rider based Elephant herd Optimization (Rider-EHO), which is obtained by integrating Rider Optimization Algorithm (ROA) and Elephant Herd Optimization (EHO). The proposed Rider-EHO-based Deep-ConvLSTM offered enhanced performance with smallest MSE of 0.003 and smallest RMSE of 0.059.