User query towards a cluster/group of video to retrieve videos automatically from available user dataset plays a significant role in developing multimedia technologies. Analyzing video for retrieval with essential features is considered as prior work. Here, input videos are viewed from UCF youtube action data for analysis. Then, foreground segmentation is performed to discover numerous small subset objects as every subset should specify the foreground video class. The segmented region is given for feature extraction. This work considers four different techniques for extracting features, namely chromatic moment, blur, color diversity, and reflection features. The extracted features are fused using Principle Component Analysis to eliminate the high dimensionality features as it may influences classification accuracy. The feature vectors are considered for combining the elements all together. Finally, classification is done with the Naïve Bayes classifier. The training and testing are performed with various combinations of data with 30:70, 40:60, 50:50, 60:40, 70:30, and 80:20, respectively. The efficiency of video retrieval is measured with metrics like accuracy, precision, recall, and F-measure. The proposed model is compared with the in-depth learning approach, where the anticipated model shows better results than that of the existing plan.