PREDIKSI PERGERAKAN ARAH HARGA SAHAM MENGGUNAKAN TREND DETERMINISTIC DATA PREPARATION DAN VARIASI ALGORITMA BOOSTING

MOCH. ANJAS A, Dr. Abdurakhman, M.Si.

2021 | Tesis | MAGISTER MATEMATIKA

Machine learning merupakan aplikasi dari disiplin ilmu kecerdasan buatan yang menggunakan teknik statistika untuk menghasilkan suatu model otomatis dari sekumpulan big data, dengan tujuan memberikan komputer kemampuan untuk "belajar". Proses belajar tersebut menggunakan algoritma khusus yang disebut algoritma machine learning. Salah satu algoritma machine learning yang paling popular dalam dunia data science adalah supervised learning. Penelitian ini bertujuan memprediksi arah pergerakan harga saham pada beberapa saham yang likuid seperti saham LQ45 dengan algoritma supervised learning seperti discrete adaptif boosting dan real adaptif boosting. Evaluasi harga saham diambil selama 10 tahun dari 2011 sampai 2021. Variabel masukkan adalah sepuluh indikator teknikal yang direpresentasikan sebagai data trend deterministic dan data nondeterministic. Proporsi data training dibagi menjadi empat yaitu 60%, 70%, 80%, dan 90% dengan melihat performa model menggunakan hasil akurasi, sensitifitas, dan spesifisitas. Berdasarkan hasil analisis dapat disimpulkan bahwa model, proporsi data training, dan jenis data berpengaruh pada hasil prediksi arah pergerakan harga saham. Terdapat model prediksi yang memberikan performa model yang sama ketika data direpresentasikan sebagai data trend deterministic dan data nondeterministic. Setiap model prediksi menunjukkan performa yang baik dengan data nondeterministic.

Machine learning is an application of artificial intelligence which uses statistical techniques to generate an automated model from collections of big data, with the aim of giving computers the ability to "learn". The learning process uses a special algorithm called a machine learning algorithm. One of the most popular machine learning algorithms in the world of data science is supervised learning. This study aims to predicted the direction of stock price movements on several liquids stocks in LQ45 with supervised learning algorithms such as discrete adaptive boosting and real adaptive boosting. The stock price evaluation is taken for 10 years from 2011 to 2021. The input variables are ten indicators technical data which is represented as deterministic trend data and nondeterministic data. The proportion of training data is divided into four, namely 60%, 70%, 80%, and 90% by looking at the performance models using accuracy, sensitivity, and specificity. Based on the results of the analysis, it can be concluded, that the model, the proportion of data training, and the type of data affect the prediction of stock price movements. There are models give the same performance when data is represented as deterministic trend and nondeterministic. All predictive models also showed well performance with nondeterministic data.

Kata Kunci : Machine Learning, Supervised Learning, Saham, Discrete Adaptive Boosting, Real Adaptive Boosting, Trend Deterministic, Nondeterministic