DETEKSI KAGGLE BOT ACCOUNT MENGGUNAKAN DEEP NEURAL NETWORKS

Abstract

Data collection in research is challenging, especially on Kaggle, a popular platform for data scientists. However, in recent years, there have been many reports of fake accounts on Kaggle that are difficult to detect, threatening data integrity and research credibility. One of the key traits to identify fake accounts is by looking at incomplete or inconsistent profiles. This research aims to help datascience users to detect Kaggle bot accounts by building a model using Deep Neural Networks. DNN is a machine learning algorithm that mimics the nervous system in the human brain. DNN consists of input layers, hidden layers, and output layers. DNN has the advantage of learning patterns from complex data and providing more accurate results than traditional Machine Learning algorithms. This research uses a dataset consisting of 1,048,574 rows of row data and 17 variables obtained from kaggle.com. The data is then pre-processed to prepare the dataset and build a DNN model with 5 hidden layers. This DNN model will be trained using training data and tested using testing data. The results show high accuracy, with 99.90% accuracy on training data, 98.42% validation accuracy, and 99.86% accuracy on testing data. These results prove that DNN can work effectively to detect fake Kaggle accounts. With this fake account detection, the quality of research can be further improved