Tatik Widiharih
Departemen Statistika, Fakultas Sains Dan Matematika, Universitas Diponegoro

Published : 48 Documents Claim Missing Document
Claim Missing Document
Check
Articles

PENERAPAN TEXT MINING DAN FUZZY C-MEANS CLUSTERING UNTUK IDENTIFIKASI KELUHAN UTAMA PELANGGAN PDAM TIRTA MOEDAL KOTA SEMARANG Genisia Pramestiloka Aulia; Tatik Widiharih; Iut Tri Utami
Jurnal Gaussian Vol 12, No 1 (2023): Jurnal Gaussian
Publisher : Department of Statistics, Faculty of Science and Mathematics, Universitas Diponegoro

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.14710/j.gauss.12.1.126-135

Abstract

Customer complaints can be handled effectively by identifying the main complaints that cause customers to be dissatisfied. Many customer complaints cause difficulty for PDAM Tirta Moedal Semarang to identify problems, which are frequently the primary complaints of customers. Grouping complaints that have similarities using Fuzzy C-Means Clustering will make the identification of the main customer complaints easier. Fuzzy C-Means uses fuzzy models, allows data to be a member of all formed clusters with membership level between 0-1. Fuzzy C-Means Clustering can also introduce more flexible patterns and show results in more accurate cluster placement. Text mining is used to convert textual data into numerical data. Customer complaints received through all contacts in PDAM Tirta Moedal Semarang from October–December 2021 were used as data. The clustering process forms 6 clusters,with the number of clusters tried being 3, 4, 5, and 6, which are seen by the smallest Xie-Beni Index. The main complaints from PDAM Tirta Moedal Semarang customer that seen through Word cloud in each cluster are that the water stops running in clusters 1 and 6 and the pipes leak in clusters 4 and 5. Complaints in clusters 2 and 3 are complaints related to water meters and water flow.
ANALISIS VOLATILITAS BITCOIN MENGGUNAKAN MODEL ARCH DAN GARCH Dheanisa Widyanti; sudarno sudarno; Tatik widiharih
Jurnal Gaussian Vol 12, No 2 (2023): Jurnal Gaussian
Publisher : Department of Statistics, Faculty of Science and Mathematics, Universitas Diponegoro

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.14710/j.gauss.12.2.254-265

Abstract

The popularity of Bitcoin increased significantly in 2021. Bitcoin is considered to deliver high returns in a relatively short period, indicating that bitcoin has high volatility. Data with high volatility usually violates the Autoregresstive IntegratedinMovinginAverage (ARIMA)in homoscedasticity assumption. The Autoregressive Conditional Heteroscedasticity (ARCH) and General Autoregressive Conditional Heteroscedasticity (GARCH) model is often used to overcome the problem of heteroscedasticity in thelARIMA model. The ARCH and GARCH models canfbe used to model thefvolatilityfof data. This Research uses ARCH and GARCH models to overcome the heteroscedasticity problem caused by the high volatility of Bitcoin data for the period 30th June 2018 to 30th June 2022. The results of this study suggest that there might be a heteroscedasticity problem in Bitcoin data. The bestffiimodel for Bitcoin data ismiARIMA(1,0,[4])-GARCH(1,1) with an AIC value of -1,4263 at a 95% confidence level
PENERAPAN KLASIFIKASI REGRESI LOGISTIK BINER DAN ADAPTIVE BOOSTING MENGGUNAKAN CLASSIFICATION AND REGRESSION TREES PADA PREDIKSI PENYAKIT HEPATITIS C Ellina Dhiya Ulhaq Oktaviani; Rukun Santoso; Tatik Widiharih
Jurnal Gaussian Vol 15, No 1 (2026): Jurnal Gaussian
Publisher : Department of Statistics, Faculty of Science and Mathematics, Universitas Diponegoro

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.14710/j.gauss.15.1.120-130

Abstract

Chronic liver disease is primarily attributed to the hepatitis C virus. Disorders of liver function can inhibit metabolism and threaten health. Hepatitis C disease must be detected earlier to reduce the risk of spreading it. Data processing using the Binary Logistic Regression and Adaptive Boosting classification methods to predict the category of patients with positive or negative hepatitis C status. Problems with unbalanced data are found in the classification process. Data imbalance can be overcome with the Synthetic Minority Over-Sampling Technique (SMOTE). Data retrieval was obtained from the 2020 UCI (University of California Irvine) Machine Learning Repository regarding data on predictions of hepatitis C patients which were downloaded on October 25, 2022. The results for the accuracy of the classification show that the Binary Logistic Regression method produces an accuracy value of 97,44%, the value sensitivity of 100%, and specificity of 97,17%. The accuracy of the classification produced by the Adaptive Boosting method with an accuracy value of 92,31%, a sensitivity value of 63,64%, and specificity of 100%. Binary Logistic Regression is the best method that can classify hepatitis C status of patients with the highest sensitivity of 100%.
KLASIFIKASI SENTIMEN KASUS ONLINE TRADING BINOMO PADA TWITTER MENGGUNAKAN ALGORITMA CONVOLUTIONAL NEURAL NETWORK (CNN) Elva Nadia Nugroho; Tatik Widiharih; Arief Rachman Hakim
Jurnal Gaussian Vol 15, No 1 (2026): Jurnal Gaussian
Publisher : Department of Statistics, Faculty of Science and Mathematics, Universitas Diponegoro

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.14710/j.gauss.15.1.46-56

Abstract

Investment in futures trading and technology is growing in Indonesia, making many domain sites of online trading companies appear that are easy for everyone to access. The Binomo app is so viral among the public because of an ad that displays professional traders who can earn 1000 USD a day without leaving their homes. Binary options trading ultimately causes losses for some people, allegedly caused by affiliates. The Binomo case is widely discussed on social media, especially Twitter. Tweets displaying public opinion on Twitter can be used for sentiment analysis and categorizing public opinion on the Binomo case. 2.686 tweets were collected by Twitter scraping between January 1 and April 1, 2022. 1.519 tweets were left after pre-processing. The data were processed using the Convolutional Neural Network algorithm with the Word2Vec method to determine their accuracy and identify topics often discussed by the public on Twitter. A CNN model with 70% training data, 3, 4, 5, kernel sizes, 4 batch sizes, 30 epochs, and Adam optimizers was used to build the classification in this research. The accuracy value obtained from the performance evaluation of the Convolutional Neural Networ model research was 89%.
ANALISIS KLASIFIKASI MENGGUNAKAN REGRESI LOGISTIK BINER DAN K-NEAREST NEIGHBOR PADA DATA IMBALANCE Eva Fitriyani; Tatik Widiharih; Bagus Arya Saputra
Jurnal Gaussian Vol 15, No 1 (2026): Jurnal Gaussian
Publisher : Department of Statistics, Faculty of Science and Mathematics, Universitas Diponegoro

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.14710/j.gauss.15.1.154-165

Abstract

Savings and Loan Cooperative or (KSP) is a cooperative that conducts its business activities only saving and borrowing. KSP members come from various different backgrounds so that they can affect their behavior in carrying out their obligations. To find out the status of current or bad customer payments, a classification process is carried out. The division of KSP customer data is carried out in the classification process into two, namely training data and test data. In the classification process, there are often cases of data imbalance, so it is necessary to handle data imbalance in training data with SMOTE and ADASYN. SMOTE and ADASYN were chosen because these methods handle imbalance data by generating data from minor classes so as not to eliminate important parts of the data. Classification was performed with Binary Logistic Regression and K-Nearest Neighbor. Binary Logistic Regression is a regression where the dependent variable is binary. While K-Nearest Neighbor is a grouping method based on the closeness of the distance of a data with other data as many as k nearest neighbors. The results of this study indicate that the ADASYN Binary Logistic Regression method is the best method that can classify and predict the payment status of KSP customers because it produces the highest accuracy and G-mean, namely the accuracy value of 70.67% and G-Mean 67.63%.
BOOTSTRAP AGGREGATING CLASSIFICATION AND REGRESSION TREES (BAGGING CART) UNTUK KLASIFIKASI POTENSI KARYAWAN RESIGN BERDASARKAN KENYAMANAN BEKERJA Muhammad Fajar Syabana; Tatik Widiharih; Masithoh Yessi Rochayani
Jurnal Gaussian Vol 14, No 2 (2025): Jurnal Gaussian
Publisher : Department of Statistics, Faculty of Science and Mathematics, Universitas Diponegoro

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.14710/j.gauss.14.2.512-523

Abstract

Phenomenon of employee resignation is a significant challenge for companies because it affect the productivity and stability of the company's operations. Every companies supposed to analyze the potential of employee resignation. This research aims to classify the potential of employee resignation based on working comfort and applies the classification modeling method from Decision Tree: Classification And Regression Trees (CART) and the ensemble Bootstrap Aggregating (Bagging) method. CART is a non-parametric method that is effective in building classification and prediction models based on decision trees, while Bagging is an ensemble method that combines several CART models to improve the accuracy and stability of predictions. The CART model provides an accuracy of 73% and f1-score of 62%, while the Bagging CART model provides an accuracy of 87% and f1-score of 88%. This research shows an increase in accuracy when using Bagging CART model of 14%. The most important variable to build the model and make predictions is the age. Age is also used as the root node in building CART model.
PENGELOMPOKAN DAERAH RAWAN KRIMINALITAS DI INDONESIA TAHUN 2021 MENGGUNAKAN METODE K-MEANS & FUZZY C-MEANS Wulan Cahya Rahma Rahmadani; Tatik Widiharih; Tarno Tarno
Jurnal Gaussian Vol 15, No 1 (2026): Jurnal Gaussian
Publisher : Department of Statistics, Faculty of Science and Mathematics, Universitas Diponegoro

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.14710/j.gauss.15.1.234-242

Abstract

Crime is one of the aspects that can influence national stability and security. In this study, crime data is used to cluster crime-prone areas and is considered whether these areas require extra surveillance or not. This research employs the K-means and Fuzzy C-Means methods. The K-means method groups data based on the similarity of data with cluster centroids. This algorithm is relatively efficient in complexity and easy to understand. K-means explicitly allocates data to specific clusters. On the other hand, Fuzzy C-Means is capable of placing a data point that lies between two or more other clusters into a single cluster. This is due to each data point having a degree of membership to determine its grouping, making the chances of failure to converge or stable cluster centers very low. The optimal number of clusters is selected using validation through the Davies Bouldin Index and Calinski Harabasz Index. The research results indicate that clustering crime-prone areas using both methods yield the same outcome of 2 clusters. Cluster 1 exhibits a higher crime rate compared to cluster 2, as indicated by the higher average values of group members. The lowest Davies Bouldin Index is 1.04948, and the highest Calinski Harabasz Index is around 24.36783.
KLASIFIKASI MENGGUNAKAN ALGORITMA K-NEAREST NEIGHBOR DAN C5.0 PADA IMBALANCE CLASS DATA DENGAN SMOTE Salsabilla Rizka Ardhana; Tatik Widiharih; Bagus Arya Saputra
Jurnal Gaussian Vol 15, No 1 (2026): Jurnal Gaussian
Publisher : Department of Statistics, Faculty of Science and Mathematics, Universitas Diponegoro

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.14710/j.gauss.15.1.280-288

Abstract

Rural Banks (BPR) provide financial services to micro-businesses and low repayment communities, especially in rural areas. The main activity of the bank is lending. Customer credit classification is expected to assist BPR in anticipating potentially bad loans. K-Nearest Neighbor and C5.0 classify current and potentially bad credit status based on customer data from BPR “X” in Central Java in October 2022. K-Nearest Neighbor is effective against a large amount of training data and works based on the nearest neighbor. C5.0 can improve classification accuracy and work by calculating entropy, gain, split info, and gain ratio to form a decision tree. There is an imbalance class data which causes the classification process to focus more on the majority class. Imbalance class data is handled using SMOTE as an oversampling approach. Classification with the addition of SMOTE can improve the evaluation of classification accuracy, especially G-Mean. G-mean is the most comprehensive measurement compared to accuracy, sensitivity and specificity in evaluating classification performance on imbalance class data. Result show an increased G-Mean to 58.55% on KNN and 64.05% on C5.0. Based on the classification results, it is concluded that C5.0 with SMOTE is a more appropriate classification model for customer credit status.