Movie Recommendation Based on Posters and Still Frames using Machine Learning

Abstract

Movie recommendation system has become a key part in online movie services to gain and maintain the huge market. While within the preceding studies works Convolution neural network (CNN) concept is employed to spot the various movies with similar posters or stills to recommend the users. Using CNN, similar posters and stills are classified into group and any hard cash within the poster may place it out of the group. But the CNN method isn't fully connected and uses backpropagation technique which could be a touch slow within the poster identification and more over just with posters the films cannot be of comparable one and should disappoint the user. Technologies like Fully Convoluted neural network (FCN) makes use of Convolution neural network concept by connecting all neural networks and adding filters and pooling layer in between each filter layer. Data Augmentation is an algorithm which helps in increasing accuracy for the predicting movies. LASSO regression is employed to get images of high multicollinearity. Soft-max layer is employed to work out the probability of the similarities int poster to create it more appropriate for the user. K-means clustering is employed to classify the films still further to recommend thes implest movietotheuser.