Vishwakarma & Thakur International Journal on Emerging Technologies 10(3): 397-403(2019) 397 International Journal on Emerging Technologies 10(3): 397-403(2019) ISSN No. (Print) : 0975-8364 ISSN No. (Online) : 2249-3255 Comparative Performance Analysis of Combined SVM-PCA for Content-based Video Classification by Utilizing Inception V3 Gagan Vishwakarma 1 and Ghanshyam Singh Thakur 2 1 Ph.D. Scholar, Department of Computer Application, Maulana Azad National Institute of Technology, Bhopal, (Madhya Pradesh), India. 2 Assistant Professor, Department of Computer Application, Maulana Azad National Institute of Technology, Bhopal, (Madhya Pradesh), India. (Corresponding author: Gagan Vishwakarma) (Received 02 August 2019, Revised 01 October 2019, Accepted 07 October 2019) (Published by Research Trend, Website: www.researchtrend.net) ABSTRACT: A multiclass classification framework based on the content of videos proposed in this paper. It is flexible to inherit standard ratings to motion pictures as class labels, prescribed by the Motion Picture Association of America (MPAA). Initially, the concept of transfer learning utilized for feature extraction using Google's inception V3 model from Data set prepared by extending the Hollywood-2 dataset and Internet Movie Database (IMDB) referred to as Extended Data Set (EDS). A modified version of the support vector machine (SVM) by combining Principal Component Analysis (PCA) projected to attain classification tasks.PCA incorporated for feature dimensionality reduction to decrease the classification complexity of multiclass SVM. Experiments illustrate a comparative analysis that the proposed, modified version combination of SVM-PCA with Inception V3 showing improved performance than classical classification algorithms like Naive Bayes (NV), Random Forest (RF), Multi-class SVM (MC-SVM). Keywords: Classification, Inception V3, Machine learning, Principal Component Analysis, Support vector machine, Transfer Learning. Abbreviations: CBFC, Central Board of film certification; MPAA, motion picture association of America; IMDB, Internet Movie Database; EDS, Extended Dataset; SVM, Support Vector Machine; PCA, Principal Component Analysis. I. INTRODUCTION At present, computer-based concepts such as machine learning [1] and recommender systems [2, 4] are very helpful in making man's work very fast [5, 6], accurate [7, 8] and efficient [2, 9]. Machine learning used for feature selection [10], Movie rating prediction [7], Emotion classification [11], and adult content detection [12] of videos or movies. Recommender system used for recommending children based movies [13] and Electronic movie recommendation [14]. The primary issue addressed in this paper is the utilization of above concepts in movie rating process by film certification agencies such as Central Board of Film Certification (CBFC) [15] belongs to India or Motion Picture Association of America (MPAA) [16, 17] belongs to united states of America. The process of movie rating entirely based on human understanding, morality, perception, and interpretations. The video or movie that is to be rated previewed in front of an examination committee formed with members of the film certification body like CBFC or MPAA. Committee formed with the condition that the number of women members not less than one-half of the total members. After the preview of the movie, every member of the examination committee submits a report in writing to the certification body for the alterations, changes, or deletions, if any. After compiling the statements following the majority view of the committee member, the certification authority issues an appropriate rating certificate for the previewed Movie or video [15]. Past researches transformed recommendation problems into less complicated rating prediction problems. Three approaches to such ratings are Collaborative, Content- based, and Hybrid recommendation [18]. In another article, a Personalized Recommendation System (PRS) based on Collaborative Filtering (CF), SVM for classification, and Improved Particle Swarm Optimization for developing personal recommendation systems proposed [14, 19]. From the above literature about the certification and rating process, we can conclude the following hypothesis: — The collaborative recommendation system is popular to develop one to five-star ratings by IMBD and Rotten tomato. — Process completed without any intervention or involvement of the computer [20, 21]. Content-based video classification is the adoption of content-based recommendation, are novel phenomena addressed in this paper. The purpose of this work is to appraise the performance of modified SVM incorporated with PCA in contrast to popular classification techniques such as Naïve Bayes, Random forests, multiclass version of classical SVM for predicting MPAA ratings based on Discrete Cosine Transform (DCT) for the attributes selection from a movie clip [16]. The rest of this paper is organized as follows in section II presents a brief background of the concepts and e t