Investigating Indonesian Payout Policy Puzzle using Data Mining Algorithm

Muasfar, Muhammad Rafi Daffa (2025) Investigating Indonesian Payout Policy Puzzle using Data Mining Algorithm. Other thesis, Institut Teknologi Sepuluh Nopember.

[thumbnail of 5031211005-Undergraduate_Thesis.pdf] Text
5031211005-Undergraduate_Thesis.pdf
Restricted to Repository staff only

Download (1MB) | Request a copy

Abstract

Meningkatnya partisipasi investor di pasar keuangan Indonesia yang sangat volatil, ditambah dengan belum adanya konsensus umum mengenai kebijakan dividen, secara teoritis semakin mempersulit urgensi penelitian. Penelitian ini bertujuan untuk mengidentifikasi model penambangan data yang paling sesuai dengan tingkat akurasi yang lebih tinggi. Variabel yang digunakan dalam penelitian ini adalah 19 prediktor, yang terdiri dari metrik keuangan, variabel spesifik tingkat perusahaan, dan kelompok makroekonomi. Sampel akhir makalah ini mencakup total 3501 observasi perusahaan-tahunan pada 651 perusahaan. Dua model pengklasifikasi utama akan digunakan dalam makalah ini, yaitu Decision Tree dan Random Forest. Temuan penelitian ini menunjukkan bahwa model Random Forest memberikan hasil yang lebih baik daripada Decision Tree dalam kinerja evaluasi model secara keseluruhan. Selain itu, hasil penelitian menunjukkan bahwa kinerja model sangat sensitif terhadap optimasi hiperparameter, sebagaimana dibuktikan oleh variasi metrik evaluasi setelah penerapan teknik regularisasi. Selain menyederhanakan perilaku keuangan perusahaan dalam mendeklarasikan dividen secara teoritis, penelitian ini bertujuan untuk membantu para pemangku kepentingan terkait (misalnya investor, manajer, dll.) dalam mengambil keputusan, dengan memberikan rekomendasi keuangan berbasis data.
===================================================================================================================================
Rising investor participants in the highly volatile Indonesian financial market coupled with no general consensus regarding dividend policy theoretically further complicated the urgency of conducting study. This study aims to identify the most suitable data mining model with higher accuracy rate. The variables incorporated in this research are 19 predictors, consisting of financial metrics, firm-level specific, and macroeconomics group. The final sample of this paper encompasses a total of 3501 firm-year observation on 651 companies. Two major classifier models will be carried out in this paper such as Decision Tree and Random Forest. The findings reveal that the Random Forest model yield better result than the Decision Tree in overall model evaluation performance. Additionally, the results demonstrate that the model's performance is highly sensitive to hyperparameter optimization, as evidenced by variations in evaluation metrics following the implementation of regularization techniques. In addition to simplify the behavioral finance of companies in declaring their dividend theoretically, this study intends to assist relevant stakeholders (e.g. investors, managers, etc.) in dealing with decisions, providing them data-driven financial recommendation.

Item Type: Thesis (Other)
Uncontrolled Keywords: Binary classification, data mining, Decision Tree, dividend, Random Forest, Klasifikasi binary, pengolahan data
Subjects: H Social Sciences > HG Finance > HG4529 Investment analysis
Divisions: Faculty of Creative Design and Digital Business (CREABIZ) > Business Management > 61205-(S1) Undergraduate Thesis
Depositing User: Muhammad Rafi Daffa Muasfar
Date Deposited: 24 Jul 2025 03:29
Last Modified: 24 Jul 2025 03:30
URI: http://repository.its.ac.id/id/eprint/120933

Actions (login required)

View Item View Item