Sholeh, Mahrus
Unknown Affiliation

Published : 2 Documents Claim Missing Document
Claim Missing Document
Check
Articles

Found 2 Documents
Search

SISTEM PERAMALAN HASIL PRODUKSI JAGUNG DI KABUPATEN SUMENEP DENGAN PENDEKATAN JARINGAN SYARAF TIRUAN BACKPROPAGATION Dafid, Ach; Sukri, Hanifudin; Sholeh, Mahrus
Jurnal Simantec Vol 12, No 2 (2024): Jurnal Simantec Juni 2024
Publisher : Universitas Trunojoyo Madura

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.21107/simantec.v12i2.26036

Abstract

Forecasting is an attempt to predict future conditions by testing past data. This forecasting is carried out on corn harvest results based on previous corn harvest data including land area, harvest area, and productivity, using the Backpropagation Artificial Neural Network forecasting system. Because the amount of corn harvest data in Sumenep Regency is very complex and changing, the backpropagation method is very suitable to be applied because it is able to handle complex and changing data. The data used in this study were collected from the book “Sumenep in Figures”. The corn production data used were from 2011 to 2023. The results of the study showed that in the scenario of varying the number of learning rates with values of 0.001, 0.2, 0.4, and 0.8, it was found that the smaller the learning rate in the Backpropagation Artificial Neural Network, the better the MSE value in the validation process. The MSE value from the results of testing learning rates of 0.001, 0.2, 0.4, and 0.8 is 0.008998. In the scenario of varying the number of iterations of 100, 500, and 1000, it is concluded that the more iterations in the Backpropagation Neural Network training, the better the MSE value in the validation process. The prediction results in the 2024 corn harvest test showed good and accurate results with a predicted value per June of 336 tons and a monthly error value of 0.0256 so that the prediction results were higher than the actual data.Keywords: ANN, Backpropagation, Forcasting System, Maize.
Peningkatan Akurasi Prediksi Harga Barang Impor Menggunakan XGBoost dan Particle Swarm Optimization Haris, Asmuni; Sholeh, Mahrus; Muflikhah, Lailil; Yudistira, Novanto
Jurnal Teknologi Informasi dan Ilmu Komputer Vol 12 No 2: April 2025
Publisher : Fakultas Ilmu Komputer, Universitas Brawijaya

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.25126/jtiik.2025129419

Abstract

Impor di Indonesia dilakukan untuk memenuhi kebutuhan dalam negeri dan memastikan kelancaran produksi serta distribusi. Namun sering terjadi under invoicing, yaitu harga barang yang diimpor dilaporkan lebih rendah dari nilai sebenarnya, yang mengakibatkan kerugian penerimaan negara. Penelitian ini bertujuan untuk memprediksi harga barang impor yang sebenarnya guna mengurangi kerugian tersebut. Data yang digunakan diperoleh dari dataset barang impor yang tersedia di platform Kaggle, yang disediakan oleh Data Analytics Community (Mof-DAC) dari Kementerian Keuangan Indonesia. Metode yang diusulkan meliputi beberapa langkah, dimulai dengan ekstraksi fitur menggunakan Large Language Model (LLM) dan Regular Expression (Regex), diikuti oleh optimasi hyperparameter XGBoost menggunakan Particle Swarm Optimization (PSO). Hasil penelitian menunjukkan bahwa model dengan ekstraksi fitur menggunakan metode Regex mengungguli LLM berdasarkan nilai Mean Squared Error (MSE), Root Mean Squared Error (RMSE), dan Mean Absolute Percentage Error (MAPE). Kombinasi ekstraksi fitur menggunakan Regex dan TFIDF memberikan hasil yang optimal dalam hal waktu pemrosesan dan akurasi prediksi. Hyperparameter terbaik untuk XGBoost ditemukan dengan max-depth 51,49, subsample 0,89, dan min_child_weight 0,65, yang meningkatkan akurasi MAPE menjadi 14,6%. Meskipun model Random Forest memiliki akurasi prediksi sedikit lebih baik dengan MAPE sebesar 12,8%, namun waktu pemrosesannya sangat lama sekitar 3 jam membuatnya kurang efisien. Sebaliknya, XGBoost dengan waktu pemrosesan hanya 51,49 detik dan MAPE 14,6% dipilih sebagai model terbaik karena akurasi yang cukup baik dengan waktu komputasi yang cepat.   Abstact Imports in Indonesia fulfill domestic needs and sustain manufacturing and distribution. Under invoicing, where imported products are purposely underpriced, reduces state revenue. This study predicts imported goods prices to reduce financial losses. The Data Analytics Community (Mof-DAC) of the Indonesian Ministry of Finance provided the Kaggle imported products dataset. The Large Language Model (LLM) and Regular Expression are used to extract features in the suggested method. XGBoost hyperparameters are then optimized using Particle Swarm Optimization. Research shows that the Regex-extracted feature model outperforms the LLM model in MSE, RMSE, and MAPE. Regex feature extraction and TFIDF produce the best processing time and prediction accuracy. The ideal XGBoost hyperparameters were a maximum depth of 51.49, a subsample value of 0.89, and a minimum child weight of 0.65. These hyperparameters increased MAPE accuracy to 14.6%. The Random Forest model has a Better Prediction Accuracy (MAPE) of 12.8%, but its processing time is 3 hours, lowering its efficiency. XGBoost was chosen as the best model due to its 51.49-second processing time and 14.6% MAPE. High accuracy and efficient computing make this model effective.