Optimasi Hyperparameter TensorFlow dengan Menggunakan Optuna di Python: Study Kasus Klasifikasi Dokumen Abstrak Skripsi

Autor: Siti Mujilahwati, Retno Wardhani, Miftahus Sholihin
Rok vydání: 2021
Předmět:
Zdroj: JURNAL MEDIA INFORMATIKA BUDIDARMA. 5:1084
ISSN: 2548-8368
2614-5278
DOI: 10.30865/mib.v5i3.3090
Popis: In today's rapidly growing digital era, the role of computing in artificial intelligence is needed to be able to help business people. Both in the fields of economy, health, and education. The use of machine learning will help related parties in viewing, analyzing, and making decisions. With machine learning, all problems related to data can be solved quickly and precisely. The problem is that the thesis document will increase every year, it will become a useless document if the data processing is not carried out. Past thesis data can be used for analysis and decision-making in the next thesis era. Python is one of the most popular programming languages used for machine learning. One reason is that there are many python-based libraries. Keras is a python-based machine learning library. TensorFlow can be used when dealing with large amounts of data processing, including thesis abstract data. Thus, this study classified 140 thesis abstract documents using hard-TensorFlow with the aim that based on the abstract content it would be classified into 6 classes, namely Android Applications, Data Mining, RPL, SPK, Digital Image Processing, and Expert Systems. The results of the classification with training data as many as 82 documents with model setting batch size = 12 and epoch = 2 with an Accuracy value of 89.04%. While the test loss test data has a higher value than the Accuracy value obtained by 66.66%. By utilizing maximizing TensorFlow performance by adding a parameter that Scikit Learn has, namely Optuna. The test data was optimized with a trial value of 500, the Accuracy increased to 76.19%
Databáze: OpenAIRE