Letter Detection : An Empirical Comparative Study of Different ML Classifier and Feature Extraction

Abstract

Work and communication activities are inextricably linked. Letters are an example of a communication medium that is still widely utilized. When it comes to significant job, however, simply an official letter is required. Official and private letters must be distinguished and classified. Different feature extraction methods, such as the count-vectorizer and TF-IDF vectorizer, are employed to transmit the detection of this official and personal letter. To categorize letters by type, various machine learning (ML) techniques are employed. Nave Bayes, Support vector machine, and AdaBoost are the algorithms. The accuracy measurements used in this study include accuracy scores, F1-mean, recall, and precision. The best working algorithm is Naïve Bayes for two vectorizer methods used, with an accuracy value of 98%.