Pemodelan dan Prediksi Poin Akhir Tim Liga 1 Indonesia Menggunakan Generalized Linear Models (GLMs)

Habibi, Agil Ammar (2026) Pemodelan dan Prediksi Poin Akhir Tim Liga 1 Indonesia Menggunakan Generalized Linear Models (GLMs). Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5003221189-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (12MB) | Request a copy

Abstract

Liga 1 Indonesia memiliki tingkat persaingan yang tinggi dengan variasi performa tim yang dinamis antarmusim, sehingga memunculkan tantangan dalam memprediksi klasemen akhir. Penelitian ini bertujuan memodelkan dan memprediksi perolehan poin akhir tim menggunakan pendekatan Generalized Linear Models (GLMs) melalui komparasi kinerja Regresi Poisson dan Regresi Binomial Negatif. Penelitian menggunakan data panel tidak seimbang sebanyak 72 observasi klub-musim yang mencakup kompetisi penuh dari musim 2021/2022 hingga 2024/2025. Hasil eksplorasi awal menunjukkan adanya indikasi marginal overdispersion dengan rasio varians terhadap rata-rata sebesar 3,77 pada data mentah. Namun, hasil komparasi model menunjukkan bahwa Regresi Poisson mampu memberikan kinerja yang lebih baik berdasarkan ukuran kesesuaian model penuh dengan menghasilkan nilai log-likelihood maksimum sebesar -215,74 dan Pseudo R-squared (Cox-Snell) mencapai 0,9707, sedangkan Regresi Binomial Negatif menghasilkan nilai log-likelihood sebesar -345,96 dan Pseudo R-squared sebesar 0,0786. Berdasarkan Uji Wald pada taraf signifikansi 5%, faktor yang berpengaruh signifikan terhadap perolehan poin akhir adalah Total Gol Cetak dan Total Gol Kebobolan. Interpretasi menggunakan Incidence Rate Ratio (IRR) menunjukkan bahwa setiap tambahan satu gol cetak meningkatkan ekspektasi poin sebesar 1,43%, sedangkan setiap tambahan satu gol kebobolan menurunkan ekspektasi poin sebesar 1,80%. Pengujian menggunakan data luar sampel (out-of-sample) pada musim Liga 1 Indonesia 2025/2026 menghasilkan nilai Mean Absolute Error (MAE) sebesar 4,00 poin dan Root Mean Square Error (RMSE) sebesar 4,68 poin, yang menunjukkan bahwa model memiliki kemampuan prediksi yang baik. Hasil penelitian ini menunjukkan bahwa Regresi Poisson efektif digunakan untuk memprediksi perolehan poin akhir tim Liga 1 Indonesia serta mengidentifikasi faktor-faktor taktis yang paling berpengaruh terhadap performa tim.
======================================================================================================================================
The Indonesian Liga 1 features a high level of competition with dynamic variations in team performance across seasons, presenting challenges in predicting the final standings. This study aims to model and predict the final points obtained by teams using the Generalized Linear Models (GLMs) framework by comparing the performance of Poisson Regression and Negative Binomial Regression. The study utilizes an unbalanced panel dataset of 72 club-season observations covering full competitions from the 2021/2022 to 2024/2025 seasons. Initial exploratory analysis indicates marginal overdispersion in the raw data, with a variance-to-mean ratio of 3.77. However, the model comparison reveals that the Poisson Regression provides superior performance based on the full model fit criteria, yielding a maximum log-likelihood of -215.74 and a Pseudo R-squared (Cox-Snell) of 0.9707, whereas the Negative Binomial Regression generates a log-likelihood of -345.96 and a Pseudo R-squared of 0.0786. Based on the Wald Test at a 5% significance level, the factors that significantly influence the final points are Total Goals Scored and Total Goals Conceded. Interpretation using the Incidence Rate Ratio (IRR) indicates that each additional goal scored increases expected points by 1.43%, while each additional goal conceded decreases expected points by 1.80%. Out-of-sample validation using external data from the 2025/2026 Liga 1 Indonesia season produces a Mean Absolute Error (MAE) of 4.00 points and a Root Mean Square Error (RMSE) of 4.68 points, demonstrating that the model possesses strong predictive capability. The results of this study show that Poisson Regression is effectively utilized to predict the final points of Liga 1 Indonesia teams and to identify the tactical factors most influential to team performance.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Liga 1 Indonesia, Generalized Linear Models, Regresi Poisson, Incidence Rate Ratio, Prediksi Out-of-Sample, Liga 1 Indonesia, Generalized Linear Models, Poisson Regression, Incidence Rate Ratio, Out-of-Sample Prediction
Subjects:	Q Science > Q Science (General) > Q180.55.M38 Mathematical models Q Science > QA Mathematics > QA278.2 Regression Analysis. Logistic regression
Divisions:	Faculty of Science and Data Analytics (SCIENTICS) > Statistics > 49201-(S1) Undergraduate Thesis
Depositing User:	Agil Ammar Habibi
Date Deposited:	30 Jul 2026 04:38
Last Modified:	30 Jul 2026 04:38
URI:	http://repository.its.ac.id/id/eprint/139644

Actions (login required)

View Item