Almehadi/AF-SOMALI-FAKE-NEWS-DETECTION-SYSTEM-

A machine learning algorithm that are correctly predicts a Somali language fake news # AF-SOMALI FAKE NEWS DETECTION SYSTEM Af-Somali language is one of under-resourced language which doesn’t have more language processing tool and techniques. One of difficulties in creating a detection technique that is reliable is lack of proper datasets and good word embedding. This project’s main objective is to design and evaluate a machine learning algorithm that are correctly predicts a Somali language fake news. The specific objectives are to review literature on fake news detection and classification methods,to develop a dataset for Somali language fake news and to evaluate the performance of the machine learning algorithms for detecting Somali fake news. # Proposed Solution The Af-Soomaali fake news detection system architecture is composed of six components namely data collection and annotations, data preprocessing, feature extraction, dataset splitting, model training and building, model testing and evaluations. First, news articles will be collected from various sources then it will be annotated and merged into a single file. Data will be preprocessed before using ML classifier using dataset cleaning, stemming, tokenizing, stop word removal, and normalization subcomponents of data preprocessing. The Word2Vec and term frequency-inverse document frequency (TF-IDF) vector space model applied to dataset to extract feature in feature extraction components. In dataset splitting component, splitting the dataset into training and test dataset conducted. Using the training dataset which contain a set of fake and real news, ML classifiers will train and learn for the dataset in model training and building components. In model testing and evaluation component, the testing dataset will be used to assess the classification performance of our models will be evaluated using accuracy, recall, precision, F1 score, and receiver operating characteristic # Results The classification result is shown in the table below, and each experiment's overall performance is qui …