The content-based video retrieval system has been the subject of 10 years of research, with an emphasis on developments, applications, and technology. In video retrieval systems, low-level materials and high-level semantic elements are still needed. Content-based video retrieval has become necessary due to the explosive growth of multimedia data caused by high-tech devices such as digital video recorders and mobile cameras. Conventional query-by-text retrieval is unable to satisfy user needs, and query-by-image retrieval is unable to match user interests with videos. By identifying temporal patterns in video information, incorporating an effective indexing and sequence matching strategy, and exhibiting encouraging efficiency and effectiveness findings, this work presents a novel approach to enhance content-based video retrieval. Four approaches are used in this study to analyze YouTube input videos: chromatic moment, blur, color variety, and reflection. Foreground segmentation and feature extraction are also performed. Feature vectors are concatenated after principle component analysis is used to eliminate high dimensionality traits. For classification, the Nave Bayes classifier is employed. A variety of metrics, including recall, accuracy, precision, and F-measure, are used to assess how successful video retrieval is. When compared to the in-depth learning method, the anticipated model performs better than the existing technique.