Journal of Dinda : Data Science, Information Technology, and Data Analytics

Analysis of Student Academic Performance to Identify New Patterns Using Linear Regression Algorithm

Adelia Putri Septiani, Akhmad Khanif Zyen, Buang Budi Wahono — Thu, 06 Feb 2025 00:00:00 +0800

Abstract

This research aims to analyze and identify new patterns in student academic performance using linear regression algorithms. Using data from 1001 respondents, this study analyzes the relationship between various variables such as study hours, previous scores, extracurricular activities, sleep hours, and learning practices on academic performance index. The research methodology employs a quantitative approach with linear regression analysis to identify relationships between variables. The results show significant correlations with an R-squared value of 0.783, indicating that 78.3% of the variation in performance index can be explained by the studied variables. Key findings reveal a synergistic effect between study hours and active learning practices, with performance improvements of up to 23%. The research also identifies a threshold effect on study hours above 6 hours which no longer provides significant impact. Optimal sleep patterns of 7-8 hours show positive correlation with highest academic performance. This study provides important contributions to understanding the factors influencing academic performance and can be used as a basis for developing more effective learning strategies.

Keywords: academic performance, linear regression, learning patterns, educational data analysis, performance index.

Integration of RFM Method and K-Means Clustering for Customer Segmentation Effectiveness

Nafissatus Zahro, Nadia Annisa Maori, Gentur Wahyu Nyipto Wibowo — Thu, 06 Feb 2025 00:00:00 +0800

Penelitian ini bertujuan untuk mengintegrasikan metode RFM dan K-Means Clustering untuk segmentasi pelanggan. Rumusan masalah yang diajukan adalah bagaimana mengintegrasikan kedua metode ini agar segmentasi pelanggan lebih efektif. Data transaksi pelanggan ADAPTA.Id tahun 2022 yang meliputi 2.252 transaksi pelanggan dianalisis untuk menghasilkan nilai RFM, dinormalisasi, dan diklaster menggunakan K-Means. Dua klaster optimal diidentifikasi dengan skor silhouette sebesar 0,8511. Dari total 2.252 transaksi pelanggan, terdapat dua klaster utama: klaster pertama berisi 10 pelanggan dengan frekuensi pembelian tinggi dan nilai transaksi signifikan, sedangkan klaster kedua terdiri dari 918 pelanggan dengan frekuensi dan nilai transaksi lebih rendah. Mayoritas pelanggan berada di klaster kedua. Segmentasi ini memungkinkan perusahaan untuk merancang strategi pemasaran yang lebih efektif dengan memfokuskan sumber daya untuk mempertahankan pelanggan bernilai tinggi dan meningkatkan aktivitas pembelian di klaster bernilai rendah. Pendekatan ini menawarkan wawasan mendalam untuk strategi bisnis yang lebih efisien, serta meningkatkan kepuasan dan loyalitas pelanggan. Skor silhouette yang tinggi menegaskan validitas klaster.

Climate Change Sentiment Analysis using LSTM

Marchel Yusuf Rumlawang Arpipi, Teny Handhayani, Janson Hendryli — Thu, 06 Feb 2025 00:00:00 +0800

This research aims to observe the sentiment of Indonesian people towards climate change using the Long Short-Term Memory (LSTM) methods. The data samples used in this study are primary data that have been collecting by using the Twitter Application Programming Interface (API) that provides by a platform known as RapidAPI. This data sample is text data with 2425 total samples obtained during the time period from 01 January 2020 to 25 August 2024. The sentiment is classified as positive, negative, and neutral. The performance of the LSTM model is evaluate using accuracy, precision, recall, F1-score, and confusion matrix and then compare with other models such as Ensemble Model, Naive Bayes, and Linear SVC. By conducting Exploratory Data Analysis (EDA), it is reveals that the distribution of public sentiment towards climate change in Indonesia from the collected data is mostly positive. However, there are not many individuals that are still ignorant and skeptical about the issue, resulting in a negative sentiment that can be fatal to the environment and its surroundings. When comparing the Ensemble Model, Naive Bayes, and Linear SVC, the LSTM model successfully identifies the perception patterns between sentences according to their sentiments. LSTM obtains an accuracy of 60% and outperforms Ensemble Model, Naive Bayes, and Linear SVC. This research also highlights the technical challenges in processing and analyzing dynamic and diverse data so that the results obtained are better, especially in terms of data quality before further processing.

Using the Random Forest Method in Predicting Stock Price Movements

Muhammad Amsari Lubis, Samsudin Samsudin — Fri, 07 Feb 2025 00:00:00 +0800

In the era of globalization, rapid technological advancements have significantly impacted the financial sector, particularly stock price movements. This study aims to contribute to financial analysis and investment by providing a predictive tool to help investors make more informed investment decisions. The Random Forest method, a machine learning al-gorithm known for effectively handling complex and heterogeneous data, is used to pre-dict stock price movements. The study utilizes historical stock data from companies listed on the Indonesia Stock Exchange (IDX) as a case study. The resulting predictive model demonstrates high accuracy, achieving 98% accuracy, with an R-squared (R²) value of 0.94 and a Mean Absolute Percentage Error (MAPE) of 0.40%. This research identifies key factors, such as Previous, High, Low, Volume, and Change, that significantly influ-ence stock price movements. The strengths of this study lie in its use of an extensive da-taset, involving 104 stock codes as examples, and its integration of interactive visualiza-tion via Streamlit to enhance data interpretation. This tool is expected to be a reliable solu-tion that provides superior predictive capabilities compared to traditional methods and supports more accurate investment analysis in the stock market.

Optimizing Search Efficiency in Ordered Data: A Hybrid Approach Using Jump Binary Search

Gabriella Youzanna Rorong, Syafrial Fachri Pane, M Amran Hakim Siregar — Sat, 08 Feb 2025 00:00:00 +0800

This research presents the development of a hybrid algorithm called Jump Binary Search (JBS), which integrates jump search and binary search techniques to improve search efficiency in sorted data distributions. JBS is designed to accelerate the search process using a jump technique to find the target block, after the block is identified, it is followed by a binary search to narrow down the search space. The results of this study show that the performance of JBS is superior compared to Jump Linear Search (JLS) when applied to non-uniform and ordered categorical data distributions. JBS only requires an execution time ranging from 0-15ms and 0-10ms, demonstrating efficiency and speed on elements consisting of 400 elements. The execution time of JBS demonstrates its efficiency compared to JLS. By minimizing unnecessary data access, JBS becomes the right solution for finding target elements in sorted data distribution.

Comprehensive Lakehouse Data Architecture Model for College Accreditation

Wed, 12 Feb 2025 16:31:53 +0800

Accreditation is an assessment activity that determines the feasibility of study programs at a university. College accreditation data comes from various sources and includes multiple data types: semi-structured, unstructured, or structured. Over time, the volume of data will continue to grow and develop, so there is a possibility of data redundancy and a long time to collect the data needed for accreditation activities. The solution is integrating data. This research aims to design a data architecture to facilitate the management of university accreditation data using the Lakehouse data architecture model. All data types can be stored on one platform in the Lakehouse data architecture. In this research, the identification, integration, and data transformation process for university accreditation data is carried out. The data used in this research is academic data in which there are with. The study's results provide an overview of the data flow process in the Lakehouse data architecture model to help better manage university accreditation data. This architecture also supports real-time data analysis so that the accreditation process can be carried out more effectively and efficiently.

Keywords: accreditation, data analysis, data architecture, data lakehouse, data warehouse

Combination of Fuzzy Weighted Product and Entropy in Determining the Eligibility of Poor Rice Recipients

Intan Zahira, Sriani Sriani — Fri, 14 Feb 2025 00:00:00 +0800

The Indonesian currency crisis caused a significant decline in food production, which ultimately resulted in increased costs and reduced purchasing power for all basic needs. Therefore, the government is trying to launch a poverty alleviation program in the form of a social assistance program for the poor in the form of rice to ease the burden on poor families and increase their access to food that is essential for life. In order to avoid problems of inconsistency in the recipients of this assistance, a decision support system was created to determine the eligibility of recipients of poor rice assistance so that it is right on target. In determining the recipients of this social assistance, evaluation criteria are needed. The evaluation criteria required are income, age, dependents, occupation and status. The researchers developed a system that combines the weighting method (Entropy) and the decision method (Fuzzy Weighted Product) to determine recipients of social assistance. Where the data is converted into fuzzy numbers first and then processed with the decision method with the final weight of the entropy weighting method. The combination method of Fuzzy Weighted Product (F-WP) and Entropy allows for optimal selection of recipients of social assistance for the poor, and an automatic selection system that can be used by administrators to streamline the process in each district. The selection results data will be available to all registered users, sorted from highest to lowest according to target ranking, to fairly determine who will receive social assistance.

Implementation of ROP and EOQ in the Stock Management Information System at Panglong Siagian Bersaudara Based on Website

Disnu Panggabean, Samsudin Samsudin — Sun, 16 Feb 2025 00:00:00 +0800

Advances in information and communication technology have encouraged the transformation from manual systems to digital-based systems, including in stock management. This research aims to develop a website-based stock management information system using the Reorder Point (ROP) and Economic Order Quantity (EOQ) methods for Panglong Siagian Bersaudara. ROP is used to determine reorder time, while EOQ calculates the optimal number of items that must be ordered to minimize total costs. The system was designed using the Rapid Application Development (RAD) method, resulting in a fast and structured design. The implementation results show that the system is able to increase stock management efficiency by reducing the risk of shortages or excess items, optimizing storage costs, and supporting real-time data-based decision making. With the integration of ROP and EOQ, this system provides a comprehensive solution for stock management at Panglong Siagian Bersaudara.

Sandal Product Inventory Prediction System Using Apriori Algorithm on Web-Based Home Industry Dlioshoes

Septia Ona Sutra, Triase Triase — Fri, 21 Feb 2025 11:19:17 +0800

Sales of sandal products at Home Industry Dlioshoes are often faced with the problem of insufficient product inventory. This research aims to build a web-based sales information system by applying the Apriori algorithm to analyze purchasing patterns and predict product inventory needs. By utilizing sales transaction data, this system can identify product combinations that distributors often buy together. By implementing the Apriori Algorithm, it can help industry owners in making decisions regarding product inventory and can predict sales in the next period, thereby reducing the risk of product excess or shortage. The research results show that the types or models of sandals that are most popular with distributors are Heels, Flat Shoes, Mules, Ballet Shoes, High Heels, Ankle Strap and Pumps. With the highest Support value of 42% and Confidence value of 71.18%.

Comparison of Sentiment for Midi Kriing and Alfagift Apps Using SVM with TF-IDF Weighting

Diva Ananda Putra, Elma Regina Nababan — Fri, 21 Feb 2025 11:21:43 +0800

The advancement of information and communication technology has impacted various aspects of life, including shopping. With increasing internet access, online shopping apps have become a primary tool for consumers. Alfa Group, a major player in the retail industry, has launched two online shopping apps, Midi Kriing and Alfagift. This study aims to compare user sentiment for these two apps based on data from Google Play Store.Using the Support Vector Machine method with TF-IDF weighting, this research analyzes 2,000 reviews from each app. The data, collected from Google Play Store, was divided into 80% for training the model and 20% for testing it. The results indicate that Midi Kriing has an overall accuracy of 87%, while Alfagift has an overall accuracy of 85%. Both apps demonstrate strong performance in sentiment detection, but Midi Kriing is slightly superior in overall accuracy. These findings provide insights into user satisfaction with the apps and can help consumers determine the best online shopping app from Alfa Group. Additionally, the results can be used by Alfa Group to enhance the services of both apps in the future.

Analysis of Bread Sales Patterns at Queen Bakery Stores Using Algorithms Fpgrowth

Muhammad Ray Pratama Sembiring, Raissa Amanda Putri — Fri, 28 Feb 2025 00:00:00 +0800

The culinary industry, particularly the bakery business, is experiencing rapid growth with increasing competition. Changes in consumer trends, the rising number of market players, and fluctuating market dynamics pose significant challenges in maintaining stable sales. Queen Bakery, a bakery located in Medan, faces issues with fluctuating monthly sales, indicating that certain products are less in demand and that more effective marketing strategies are needed. To address this issue, the utilization of technology in data analysis is essential, particularly through the implementation of data mining techniques. Data mining enables the identification of consumer purchasing patterns more accurately and data-driven. One of the most effective algorithms for sales pattern analysis is FPGrowth, which can identify frequently occurring itemsets in transactions. Unlike the Apriori algorithm, which requires extensive computations, FPGrowth is more efficient in discovering product associations frequently purchased together. This study aims to analyze sales patterns at Queen Bakery using the FPGrowth algorithm to provide strategic insights into inventory management and product marketing. The results of this research are expected to assist Queen Bakery in improving operational efficiency, optimizing product offerings, and formulating more competitive business strategies. By implementing data mining, the bakery can gain a deeper understanding of consumer preferences, ultimately enhancing sales performance and competitiveness in the dynamic culinary market.