Peramalan Spatio-Temporal Lalu Lintas Internet Pada Jaringan Seluler Dengan Seleksi Fitur Dan Deep Learning

Setyadi, Agung Teguh (2021) Peramalan Spatio-Temporal Lalu Lintas Internet Pada Jaringan Seluler Dengan Seleksi Fitur Dan Deep Learning. Masters thesis, Institut Teknologi Sepuluh Nopember.

Text 05111950010023-Master_Thesis.pdf - Accepted Version Restricted to Repository staff only until 1 April 2023. Download (3MB)

Abstract

Banyaknya akan permintaan koneksi internet pada perangkat mobile membuat operator seluler harus menjaga dan meningkatkan kualitas layanan (Quality of Service) jaringan. Hal ini dapat diatasi dengan melakukan peramalan lalu lintas internet pada jaringan seluler yang hasilnya dapat digunakan sebagai acuan untuk pengambilan keputusan dalam pengolahan sumber daya. Oleh karena itu, hasil akurasi dari peramalan sangat penting. Penambahan korelasi spatial atau biasa disebut peramalan spatio-temporal dalam peramalan jaringan seluler dapat meningkatkan akurasi. Namun, tidak semua fitur data relevan dalam peramalan spatio-temporal. Masalah ini dapat mempengaruhi akurasi peramalan.
Penelitian ini bertujuan untuk memberikan solusi atas permasalahan pada peramalan spatio-temporal lalu lintas internet pada jaringan seluler dengan mengusulkan proses peramalan spatio-temporal dengan melakukan seleksi dan ekstraksi fitur input yang relevan dengan metode Detrended Partial Cross-Correlation Analysis (DPCCA) dan mengembangkan model peramalan dengan pendekatan deep learning Long Short-Term Memory (LSTM).
Penelitian ini diuji pada dataset lalu lintas jaringan seluler telecom Milan. Berdasarkan hasil uji coba data testing, menunjukkan model yang diusulkan mendapatkan nilai RMSE terkecil dan ukuran kinerja R2 terbesar yaitu 195,309 dan 0,912 untuk sampel grid berdasarkan jarak, 411,177 dan 0,957 untuk sampel grid berdasarkan korelasi, 10,177 dan 0,787 untuk sampel random grid.
=====================================================================================================
The large number of requests for internet connection on mobile devices makes mobile operators have to maintain and improve the quality of service (Quality of Service) networks. This can be overcome by forecasting internet traffic on cellular networks, the results of which can be used as a reference for decision making in resource processing, Hence, the accuracy of forecasting is very important. The addition of spatial correlation or so-called spatio-temporal forecasting in cellular network forecasting can improve accuracy. However, not all data features are relevant in spatio-temporal forecasting. This problem can affect the accuracy of forecasting.
This study aims to provide solutions to problems in spatio-temporal forecasting of internet traffic on cellular networks by proposing a spatio-temporal forecasting process by selecting and extracting relevant input features with the Detrended Partial Cross-Correlation Analysis (DPCCA) method. and developing a forecasting model using a deep learning Long Short-Term Memory (LSTM) approach.
This research was tested on Milan telecom mobile network traffic dataset. Based on the experimental results of data testing, it shows that the proposed model gets the smallest RMSE value and the largest performance measure R2, namely 195.309 and 0.912 for the grid sample based on distance, 411.177 and 0.957 for the grid sample based on correlation, 10.177 and 0.787 for the random grid sample.

Item Type:	Thesis (Masters)
Uncontrolled Keywords:	peramalan, spatio-temporal, seleksi fitur, deep learning, Long Short-Term Memory, forecasting, spatio-temporal, feature selection, deep learning, Long Short-Term Memory.
Subjects:	H Social Sciences > H Social Sciences (General) > H61.4 Forecasting in the social sciences H Social Sciences > HA Statistics > HA30.3 Time-series analysis Q Science > Q Science (General) > Q325.5 Machine learning. Support vector machines. Q Science > QA Mathematics > QA76.87 Neural networks (Computer Science)
Divisions:	Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Informatics Engineering > 55101-(S2) Master Thesis
Depositing User:	Agung Teguh Setyadi
Date Deposited:	28 Feb 2021 21:17
Last Modified:	28 Feb 2021 21:17
URI:	http://repository.its.ac.id/id/eprint/82984

Actions (login required)

View Item