Exploring Multiple Genres Text Classification: Classifying 61 genres of Mobile App Description based on Naïve Bayes and Count Vectorizer

Text classifying is a classic NLP problem. In our following research, we aimed at a multiple classification problem. We used the list of descriptions of a vast number of apps as a data set. and applied several different machine learning algorithms to analyze the data. Firstly, we did the preprocess of the data to mark the genres, there are 59 genres in the dataset in total. Then, we vectorized the data with both Count Vectorizer and TfidfVectorizer and use the vectorized dataset to train models through Naïve Bayes, SVM, Random Forest and KNN. Then, we did some optimization for the parameters and determined Naïve Bayes the algorithm with best accuracy and performance for this problem. Finally, we further optimized the parameter for the Naive Bayes as well as reduced the redundant genres in our dataset and get the accuracy of about 60.276%.