Performance Analysis of Terrain Classifiers Using Different Packages

The 20-newsgroup dataset includes about 20,000 articles that have been systematically classified into 20 newsgroups. The compilation of 20 newsgroups is emerging as a common source of data for studying the performance of machine learning techniques language implementations, such as language detection and word clustering. Each record in the dataset is actually a text file (a series of words). The data is trained over different classifiers such as Support Vector Machines, Naïve Bayes Model, Random Forests, Stochastic Gradient Descent, and Logistic Regression. We have used Python Jupyter notebook for code execution and packages from libraries like scikit learn (Sklearn), Natural Language Toolkit (NLTK), regular expression (RE), Language Identification (LangID), etc., to test the model parameters and accuracy measurements. By tuning the parameters and relevant initialization, we aim to test the outcomes to find out the best model with maximum accuracy.