DEVELOPMENT OF SUBSYSTEM FRONTEND WEB DASHBOARD FOR PREDICTION OF THROUGHPUT DOWNLINK 4G LTE PERFORMANCE BASED MACHINE LEARNING

• Main Author: Taufiq Rafiandi, Muhammad
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/64071
• Institution: Institut Teknologi Bandung

According data from Ericsson, LTE will remain the dominant mobile access technology based on the number of subscriptions with an increase of almost 4.5 billion mobile subscription users. This increase in number has resulted in mobile users being able to access high-speed internet services such as online gaming and video streaming. Through this problem, an idea was developed to develop a machine learning prediction method for variations in 4G LTE downlink throughput by analyzing different network quality parameters at various times and locations and investigating the main factors that affect network performance and quality. The solution will be built through the development of a web dashboard application that aims to be able to display the results of the correlation between throughput and network parameters, throughput against time and the results of evaluating performance metrics for each type of machine learning algorithm. The development of this product requires a process from the backend subsystem to be implemented into the development of this web dashboard. The method of working on this final project is to study literature, design and develop system designs, implement systems and test systems. The process of working on the product begins by developing a data loading feature to accept input data uploads and existing data, then making feature selection as a feature to be able to select a number of features that are used for analysis of throughput prediction results based on observations of correlations between throughput and other network parameters, then it will be developed throughput prediction visualization results using a variety of machine learning model types, train size dataset settings, number estimators. And the results of evaluating performance metrics for each type of machine learning. Testing this system is done by checking the functionality side of whether each display process can run according to the system design design and for the product side using a questionnaire survey to assess system performance from appearance perception to product convenience and usefulness. And the results of this study indicate that the machine learning system can be implemented and runs on the web dashboard application in functionality and product-wise, this system has been tested in functionality, can be easily used by users and adding more benefits for users in using this web dashboard system.