Abstract:
Document classification is a problem which is age-old in information retrieval. To manage text effectively and to manage large volumes of unstructured information for variety of applications document classification plays an important role. Automatic document classification makes much ease in finding the relevant information at right place and much faster by routing correct documents to the users. The proposed approach comprises of the classification of the researched data being carried out in 5G and Massive MIMO. The basic idea is to classify all the work material and research study of different types and field separately. The need of faster data speeds with more reliable services for mobile users increases rapidly, which will be further fulfilled by the future/next generation mobile networks-5G. Today 4G stations with MIMO is handling dozen of cellular traffic which will be further enhanced in 5G and Massive MIMO for future generation. A huge amount of work has been done for the efficiency in transmitting data to the certain number of users per second for different services fields like channel estimation , 5G Networks, Massive MIMO, OFDM , beam-forming for instance, that results in the existence of diverse services and research in different fields. Research for the next generation is still in progress in all the respective fields. In this thesis, we have prepared a data that includes journals, conference and research papers being published in the area of 5G technology. The date is labeled into 7 different classes by an expert in 5G communication. We then proposed an automatic methodology for the Massive MIMO and 5G document classification by creating an algorithm using Natural Language Processing (NLP) techniques, which includes tokenization, stemmer and TF-IDF. For further writing and to test the classifier I have used WEKA tool along with NLP techniques and the classification of data is performed by using Sequential Minimum Optimization (SMO) and Support Vector Machine (SVM) Classifier. Evaluation of classifier was further done via Precision, Recall and F-Measure matrixes. The results shows that our methodology gives better results that the previous methodologies. With this strategy an automated system for classifying 5G and Massive MIMO related researched work is proposed which will helps in finding and classifying the researched data material of the same type effectively.
