Comparative Study of Machine Learning and Holt-Winters Exponential Smoothing Models for Prediction of CPI's Seasonal Data

Q: Which CPI dataset and period are used in the study?

The study uses univariate CPI data for education goods and services in Malang City for 1996–2013, sourced from BPS’s “Malang City in Figures” (published 1997–2014).

Q: Which prediction models are compared for CPI forecasting?

The research compares Holt-Winters Exponential Smoothing with machine learning methods: LSTM, ELM, Ridge, and LASSO.

Q: What model performs best according to the reported error metrics?

Ridge produces the smallest Mean Absolute Percentage Error (MAPE), reported as 2.10723%, outperforming the other machine learning methods and the Multiplicative Holt-Winters Exponential Smoothing approach.

Aug 3, 2026

Upload by Eliana

Inflation influences price stability and consumers’ purchasing power, making reliable forecasting essential. This research compares CPI prediction performance for education goods and services using Holt-Winters Exponential Smoothing and machine learning models, including LSTM, ELM, Ridge, and LASSO. Univariate CPI data for Education Goods and Services in Malang City from 1996–2013 is sourced from BPS “Malang City in Figures” (1997–2014). Results show Ridge achieves the lowest MAPE (2.10723%) and all simulated machine learning models reach MAPE < 10%, indicating strong seasonal time-series predictive accuracy.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages5

Reading Time13 min

AccessFree to download

Introduction

Time-series, trends, and seasonality

Inflation and CPI as forecasting targets

Forecasting methods: smoothing and machine learning

Comparative Study of Machine Learning and HoltWinters Exponential Smoothing Models for Prediction ofCPI’s Seasonal Data 1stAshri Shabrina Afrah Department of Informatics Engineering Universitas Islam Negeri Maulana Malik Ibrahim Malang Malang, Indonesia [ashri.shabrina@ti.uin-malang.ac.id](ashri.shabrina@ti.uin-malang.ac.id) 4thKhadijah Fahmi Hayati Holle Department of Informatics Engineering Universitas Islam Negeri Maulana Malik Ibrahim Malang Malang, Indonesia [khadijah.holle@uin-malang.ac.id](khadijah.holle@uin-malang.ac.id) 2ndNur Fitriyah Ayu Tunjung Sari Department of Informatics Engineering Universitas Islam Negeri Maulana Malik Ibrahim Malang Malang, Indonesia[nur.fitriyah@ti.uin-malang.ac.id](nur.fitriyah@ti.uin-malang.ac.id) 5thMerinda Lestandy Department ofElectrical Engineering Universitas Muhammadiyah Malang Malang, Indonesia [merindalestandy@umm.ac.id](merindalestandy@umm.ac.id) 7thRizdania Computer Science Department University ofPGRI Wiranegara Pasuruan, Indonesia [rizdania.uniwara@polinema.ac.id](rizdania.uniwara@polinema.ac.id) 3rdShoffin Nahwa Utama Department of Informatics Engineering Universitas Islam Negeri Maulana Malik Ibrahim Malang Malang, Indonesia [shoffin@uin-malang.ac.id](shoffin@uin-malang.ac.id) 6thEndah Septa Sintiya Information Technology Department State Polytechnic of Malang Malang, Indonesia [e.septa@polinema.ac.id](e.septa@polinema.ac.id) Abstract— Inflation is one of the factors influencing price stability. Inflation affects people's purchasing power and impacts their decisions as economic actors. Consumer Price Index (CPI) is one of the factors used by economists to measure the inflation or deflation in a country. This research focuses on comparing the prediction results of the CPI for the Education Goods and Services using the Holt's Winters Exponential Smoothing and Machine Learning Methods, namely Long Short-Term Memory (LSTM), Extreme Learning Machine (ELM), Ridge, and Least Absolute Shrinkage and Selection Operator (LASSO). The data used is univariate data on the CPI for the Education Goods and Services in Malang City in 1996-2013, which was obtained from the publication of the Statistics Indonesia (BPS), entitled "Malang City in Figures"which was published in 1997-2014. The results of this research show that the Ridge Method produces the smallest Mean Absolute Percentage Error (MAPE) value compared to other Machine Learning Methods and the Multiplicative HoltWinters Exponential Smoothing Method, with MAPE=2.10723%%. Machine Learning models that have been simulated have very good accuracy values, with MAPE values < 10%. Therefore, it can be assumed that the simulated Machine Learning models can make very good predictions on time-series data with seasonal patterns. Keywords—machine learning, data mining, time series, seasonality, consumer price index I. INTRODUCTION Historical data usually consist of series of observed data collected within particular time interval. The series are known as ‘time-series’ [1] . Time series data is usually ordered in time and time-dependent with certain trends or seasonal patterns. The data trends and patterns can be analyzed to predict the occurrence in the future. Time series data may have fluctuations or patterns which are repeated with fixed frequencies. This is called seasonality of time series data. Inflation is one of the factors that influences price stability. According to the Statistics Indonesia (BPS), inflation is a tendency to increase the prices of goods and services in general which occurs continuously [2] . Inflation also affects people's purchasing power and impacts their decisions as economic actors. CPI is one of the factors used by economists to calculate the inflation rates. The CPI is an index number that measures the price of goods and services that are always used by consumers or households [3] . This isone of the indicators used to see monetary success in controlling inflation. The value i

FAQs

Which CPI dataset and period are used in the study?

The study uses univariate CPI data for education goods and services in Malang City for 1996–2013, sourced from BPS’s “Malang City in Figures” (published 1997–2014).

Which prediction models are compared for CPI forecasting?

The research compares Holt-Winters Exponential Smoothing with machine learning methods: LSTM, ELM, Ridge, and LASSO.

What model performs best according to the reported error metrics?

Ridge produces the smallest Mean Absolute Percentage Error (MAPE), reported as 2.10723%, outperforming the other machine learning methods and the Multiplicative Holt-Winters Exponential Smoothing approach.

Aug 3, 2026

Upload by Eliana