Knowbase : International Journal of Knowledge in Database

Application of Data Mining for Ceramic Sales Data Association Using Apriori Algorithm

M. Ilham Habibi — 2024-12-05

This research is conducted to provide an understanding of consumer purchasing patterns at CV. Sukses Bersama by applying data mining using the association rules method and the Apriori algorithm to identify the relationships between one item that influences other items within a ceramic sales dataset at CV. Sukses Bersama. This information is expected to serve as a foundation for improving sales strategies, optimizing customer satisfaction, and expanding the company's market share. The Apriori algorithm is a popular algorithm implemented to identify association rules in data mining. The Apriori algorithm was chosen due to its ability to efficiently identify association rules and its good scalability in handling large datasets. This research begins with the collection of ceramic sales data, followed by data preprocessing to clean and prepare the data. The Apriori algorithm is then applied to discover the association rules, which generate two matrices: support and confidence, and the results are subsequently evaluated. This research was conducted using Google Colaboratory, a web application that is a cloud-based platform provided by Google to run Python code. The results of the study show that the Apriori algorithm can depict significant association structures between different ceramic brand types in the sales data of CV. Sukses Bersama. The calculation results show that the rule has the maximum support and confidence value, namely 67% support value and 84% confidence value in the rule "if you buy the DIAMD brand, you will buy the TOTAL brand"

Modelling Time Series Data for Stock Prices Prediction Using Bidirectional Long Short-Term Memory

Yenie Syukriyah — 2024-12-24

The dynamic nature of stock markets, characterized by intricate patterns and sudden fluctuations, poses significant challenges to accurate price prediction. Traditional analytical methods are often unable to capture this complexity. This requires the use of advanced techniques capable of modelling non-linear dependencies. This study aims to build a model using recurrent neural network and predict the Indonesian stock prices. PT Gudang Garam Tbk.'s (GGRM.JK) stock was selected due to its significant role in the Indonesian stock market and its contribution to national revenue through excise tax. The method used in this research involves training the BiLSTM (Bidirectional Long Short-Term Memory) model using historical stock price data with training and test data ratios of 90:10, 80:20 and 70:30 to determine the optimal configuration. The evaluation results showed that the 90:10 data ratio gave the best performance with a MAPE of 1.51%, MAE of 343.55 IDR and RMSE of 522.30 IDR. These results indicate that the BiLSTM model has high accuracy and minimal prediction errors. Further analysis showed that the model performed optimally with a batch size of 32 and higher epochs, such as 200 and 250, providing greater stability and prediction accuracy. These results demonstrate the potential of the BiLSTM model as an effective predictive tool to support strategic investment decisions, particularly for high volatility stocks. Future research is recommended to test this model on other stock data and to consider external factors to improve its generalizability.

Analysis of A Priori Algorithm in Medical Data for Heart Disease Identification with Association Rule Mining

Davip Sutejo — 2024-12-24

Heart disease is one of the leading causes of death worldwide, so it is important to identify risk factors that can contribute to the development of this disease in order to carry out early prevention. This study aims to identify patterns of association between risk variables and the incidence of heart disease using the Association Rule Mining (ARM) method combined with the A priori algorithm. The data used in this study includes lifestyle information, medical history, and other health parameters, obtained from the UCI Machine Learning repository. The analysis results showed that with a support value between 30% and 70%, the strongest association rule was found between sex (sex = 1) and angina (exang = 1), with a lift value of 1.67, indicating a strong positive relationship towards a positive diagnosis (target = 1). In addition, other moderate association rules were found, such as the combination of cp_1 = 1 and ca_0 = 1, with a lift value of about 0.73, indicating a weaker association. These findings suggest that some attribute combinations have higher predictive power, which can be used to improve prediction models in the medical diagnosis of heart disease. This research also highlights the main challenges faced by the A priori algorithm, such as computational complexity and selecting the right threshold to obtain significant rules

Business Intelligence Dashboard Human Resource Capacity to Increase the Capacity City of Bekasi

R Wisnu Prio Pamungkas — 2024-12-24

Bekasi City with qualified and evenly distributed human resources will be better able to meet dynamic and complex development needs. Effective data visualization can simplify complex information related to HR capacity, such as education levels, skills distribution, and the number of workers in various sectors, making it easier for policy makers to design strategies including identifying the distribution of filling several positions based on gender and identifying areas of need for educational facilities, children's health, and other infrastructure that supports the growth and development of the younger generation, and developing more effective policies to improve the overall capacity of the city. This research aims to develop a human resource capacity data visualization model as a tool in improving city capacity. This research uses Google Looker Studio as a data visualization platform, data integration is done by Extract, Transform, Load (ETL) method, the data starts from Excel then cleaned, adjusted the format and loaded into Google Sheets. The data used includes key variables that describe the characteristics of human resources in the Bekasi city area, such as education, age group, gender, and demographic distribution. The results show that based on the dashboard visualization, the Bekasi City government can increase 10% representation of the number of women in supervisory and administrator positions in 2 years and the number of only 5% at the S2 or S3 education level requires an increase in education to support the optimization of HR for strategic positions

Design and Development of an Online Analytical Processing (OLAP) Application for Customer Profiling Analysis of Insurance "X"

Debleng Puja Kesuma Wardanie — 2024-12-31

The system's slow and inflexible response time is a characteristic of analytical processes based on transactional databases (OLTP), as experienced by PT Asuransi "X." This limitation arises because transactional databases are not designed for OLAP, which can provide various functions to perform synthesis and analysis that improve response time. This study aims to design and develop an Online Analytical Processing (OLAP) application to be used for customer profiling analysis at insurance company "X." In the insurance industry, effective and efficient data analysis is essential to understand customer behavior, perform segmentation, and make more informed decisions in marketing insurance products. The OLAP application developed in this study integrates various customer data dimensions, such as demographics, claim history, and owned products, facilitating multidimensional analysis for its users. The application design process includes system design, data collection, OLAP technology implementation, and application testing. The study results indicate that the application reveals that the majority of customers are male (56%), aged between 30 and 45 years (45%), and employed in the private sector. Additionally, in the city of Surabaya, there is a higher tendency to purchase the Mitra Sakinah life insurance policy. This information enables the company to better understand customer demographic characteristics and tailor its marketing strategies accordingly.

Optimization Of Agricultural Production In South Sumatera Using Multiple Linear Regression Algorithm

Dedi Setiadi — 2024-12-31

Rice is one of the agricultural commodities in South Sumatra whose productivity level still fluctuates. In 2000, rice production reached 1,863,643.00 kg, then increased to 3,272,451.00 kg, in 2010, but decreased again in 2020 to 2,696,877.46 kg. This instability is influenced by various factors such as land area, rainfall, pest attacks, and fertilizer use. This study aims to optimize rice production by applying machine learning using multiple linear regression algorithms, and the CRISP-DM method, with the stages being business understanding, data understanding, data preparation, modeling, evaluation, and implementation. Data of 1,000 records obtained from farmers were analyzed using Google Collaboratory, resulting in an intercept of -3836,2639, and coefficients for land area of 5,7336, rainfall of 1,2710, pests of 6,1153, urea of 1,6226, and phonska of 1,2581. To evaluate the accuracy of rice production predictions based on these independent variables, calculations were made on the RMSE value and analysis of the coefficient of determination. The results were that the RMSE value was recorded at 17065084,9641, and the coefficient of determination (R²) was 0,6487, indicating that around 64,87 % of the variability in rice production can be explained by independent variables such as land area, rainfall, pest attacks, use of urea fertilizer, and phonska, while the remaining 35,13 % was influenced by other factors.

Performance Comparison of Naïve Bayes and SVM Algorithms in Sentiment Analysis on JKN Application Data

Meyti Eka Apriyani — 2024-12-31

In 2022, 67.88% of Indonesia's population owned mobile devices. BPJS Kesehatan responded to this trend by launching the Mobile JKN application to provide modern, accessible healthcare services. To drive continuous innovation, BPJS Kesehatan needs insights into user feedback regarding the Mobile JKN application. Given the large volume of reviews, sentiment analysis is employed to classify reviews into positive or negative categories. This study compares the performance of Naïve Bayes and SVM (Support Vector Machine) algorithms in sentiment classification using a dataset from the Mobile JKN application. The dataset consists of 200 reviews labeled by two different raters, yielding 110 positive and 90 negative reviews for the first set and 114 positive and 86 negative reviews for the second set. Testing was conducted using three data split scenarios for training and testing: 70:30, 80:20, and 90:10. Model performance was evaluated using a confusion matrix, with metrics including accuracy, precision, recall, and F1-score. The results show that the Naïve Bayes algorithm achieved its best performance with a 90:10 data split, yielding an accuracy of 85%, precision of 77%, recall of 100%, and F1-score of 87%. Conversely, the SVM algorithm performed best with an 80:20 data split, achieving 93% accuracy, 100% precision, 84% recall, and an F1-score of 91% for the first rater's dataset. For the second rater's dataset, SVM reached optimal performance with a 90:10 data split, yielding 90% accuracy, 100% precision, 80% recall, and an F1-score of 89%. Overall, the comparison highlights that SVM outperforms Naïve Bayes in terms of accuracy and precision, making it more effective for predicting positive sentiment in Mobile JKN application reviews.

Enhancing Stroke Diagnosis with Machine Learning and SHAP-Based Explainable AI Models

Galih Hendro Martono — 2024-12-31

Stroke is a serious illness that needs to be treated quickly to enhance patient outcome. Machine Learning (ML) offers promising potential for automated stroke detection through precise neuroimaging analysis. Although existing research has explored ML applications in stroke medicine, challenges remain, such as validation concerns and limitations within available datasets. The study aims to compare ML models and SHapley Additive exPlanations (SHAP) algorithm insights for stroke detection optimization. The research evaluates classifiers' performance, including Deep Neural Networks (DNN), AdaBoost, Support Vector Machines (SVM), and XGBoost, using data from www.kaggle.com. Results demonstrate XGBoost's superior performance across various data splits, emphasizing its effectiveness for stroke prediction. Utilizing SHAP provides deeper insights into stroke risk factors, facilitating comprehensive risk assessment. Overall, the study contributes to advancing stroke detection methodologies and highlights ML's role in enhancing clinical practice in stroke medicine. Further research could explore additional datasets and advanced ML algorithms to enhance prediction accuracy and preventive measures.