Electronic Theses and Dissertation

Pergerakan harga saham bersifat kompleks, dinamis, dan nonlinier sehingga memerlukan metode prediksi dengan tingkat akurasi tinggi. penelitian ini bertujuan untuk mengetahui kemampuan model adaptive neuro fuzzy inference system (anfis) dalam memprediksi harga penutupan saham pt bank rakyat indonesia (persero) tbk (bbri) berdasarkan data historis harian periode juli 2015 hingga juli 2025 menggunakan variabel open, high, dan low sebagai input, serta menghasilkan gambaran pergerakan harga penutupan saham bbri untuk satu minggu ke depan menggunakan pendekatan multi-step forecasting. data yang digunakan merupakan data sekunder harga saham harian bbri sebanyak 2.410 data. tahapan penelitian meliputi normalisasi data, clustering menggunakan fuzzy c-means (fcm) dengan 3 cluster, pembentukan fungsi keanggotaan gaussian, pembagian data training dan testing dengan proporsi 80:20, serta pelatihan model anfis menggunakan algoritma hybrid learning selama 100 epoch. hasil evaluasi menunjukkan nilai mape training sebesar 2,345% dan mape testing sebesar 1,919%, didukung nilai rmse dan mae testing masing-masing sebesar 0,004 dan 0,003, yang seluruhnya termasuk kategori akurasi sangat baik. mape testing yang lebih rendah dari training mengindikasikan model tidak mengalami overfitting dan mampu melakukan generalisasi dengan baik. simulasi multi-step forecasting menghasilkan prediksi harga penutupan bbri selama 5 hari perdagangan pada kisaran rp3.924 hingga rp4.144, menggambarkan tren harga yang cenderung meningkat dalam jangka pendek. dengan demikian, model anfis efektif digunakan dalam memprediksi harga penutupan saham bbri.

Stock price movements are complex, dynamic, and non-linear, thus requiring prediction methods with high accuracy. This study aims to evaluate the performance of the Adaptive Neuro Fuzzy Inference System (ANFIS) model in predicting the closing stock price of PT Bank Rakyat Indonesia (Persero) Tbk (BBRI) based on daily historical data from July 2015 to July 2025 using Open, High, and Low prices as input variables, as well as to generate a short-term price movement forecast for the following week using a multi-step forecasting approach. The data used consists of 2,410 daily stock price observations of BBRI. The research stages include data normalization, clustering using the Fuzzy C-Means (FCM) method with 3 clusters, construction of Gaussian membership functions, data partitioning into training and testing sets with an 80:20 ratio, and training the ANFIS model using a hybrid learning algorithm for 100 epochs. The evaluation results show a training MAPE of 2.345% and a testing MAPE of 1.919%, supported by testing RMSE and MAE values of 0.004 and 0.003 respectively, all of which fall into the very high accuracy category. The testing MAPE being lower than the training MAPE indicates that the model does not suffer from overfitting and is capable of generalizing well to unseen data. The multi-step forecasting simulation produced closing price predictions for BBRI over 5 trading days ranging from Rp3,924 to Rp4,144, reflecting a gradually increasing price trend in the short term. Therefore, the ANFIS model is effective in predicting the closing stock price of BBRI.