Analysis of Stock Market Value Prediction using Novel Long Short-Term Memory in Comparison with SVM for Increased Accuracy

The proposed algorithm Novel Long Short-Term Memory (NLSTM) has the aim to improve the stock market performance prediction in the aspect of improving the accuracy, for the analysis was compared with Support Vector Machine (SVM). The estimated sample size of $\boldsymbol{\mathrm{N}=1000}$ was calculated using g-power analysis tool for each group with $\mathbf{alpha=0.05},\mathbf{power=0.8}$ and confidence $\mathbf{interval=95\%.}$ The proposed NLSTM was dependent on time collection statistics in inventory rate prediction based on time collection records, their lags with various time lapse. The model was implemented on top of ANN, along with various nodes in the network for stock market prediction. The NLSTM (63.10% has a better accuracy rate compared to SVM (48.91%) of machine learning. The statistical result shows that p=0.024 (p<0.05), which helps to decide the comparative analysis of NLSTM and SVM is significantly different. Based on the achieved accuracy and significance, evidence concludes that NLSTM has improved accuracy and is better than SVM.