In this paper, the problem of existing cache replacement and invalidation policies are examined from different dimensions namely valid scope space optimization, prediction functions, access methods, and uncertainty. The proposed Predicted Region Enrooted Method for Invalidation Efficient cache Replacement (PREMIER) policy first achieves time-series data by preprocessing the user’s movement non-stationary trajectory data and then it applies Convolutional Neural Network-Long Short-Term Memory (CNN-LSTM) Network to find out the pattern that appears frequently. Using predicted next probable location, the PREMIER approach uses a revised data item cost function for cache replacement and the CELP function for cache invalidation. The predicted region computation-based location-dependent data cache invalidation and replacement approach PREMIER achieve significant improvement in the cache hit rate efficacy as compared to that of past cache invalidation-replacement policies such as CEMP-IR, SPMC-CRP+CEB, PPRRP+ CEB, and Manhattan+CEB for LBS.