Jurnal KomtekInfo

Analysis of the feasibility level of IT device using K-Means cluster and C4.5 classification

Fachriqi Naldes — 2026-03-30

The availability of reliable laptops is essential for ensuring smooth business operations; however, decisions regarding device upgrades and replacements in many organizations still rely primarily on device age and subjective user perceptions. This practice often leads to inconsistent IT asset lifecycle decisions, increased security risks, and inefficient cost management. This study proposes a classification model to recommend laptop feasibility levels, namely usable, requires upgrade, and requires replacement, based on a combination of technical specifications and operating system characteristics. K-Means clustering is applied to group laptops into three feasibility categories using processor type, release year, RAM capacity, storage type, and operating system attributes that have undergone performance score–based ordinal encoding and Min–Max normalization. Subsequently, the C4.5 algorithm is employed to construct a decision tree using the K-Means cluster labels as target classes, producing interpretable if–then rules that describe device feasibility patterns. The dataset is obtained from the IT device inventory of PT Semen Indonesia, consisting of 1,905 laptop records, which after data cleaning result in 85 unique specification combinations for analysis. The clustering process classifies 47 laptops as usable, 22 as requiring upgrades, and 16 as requiring replacement. The C4.5 algorithm model achieves accuracy, precision, recall, and F1-score values of 100% on the test data, indicating its ability to effectively replicate the feasibility patterns generated by K-Means algorithm. These findings demonstrate that the proposed approach provides a data-driven framework for supporting upgrade and replacement decisions, contributing to more efficient and measurable IT asset lifecycle management.

Analysis of Clean Water Consumption Segmentation And Classification Using K-Means Clustering And Random Forest Algorithms

Ika Melinia Sapitri Fitriyanti — 2026-03-30

The administrative grouping of PERUMDA Air Minum Kota Padang customers is not yet able to accurately represent actual customer water consumption patterns. This condition makes it difficult for the company to formulate service policies, customer management, and make appropriate data-based decisions. This study aims to analyze and map customer water consumption patterns to produce more representative customer segmentation as a basis for decision making. The research method used is a data mining approach with the application of Principal Component Analysis (PCA) for dimension reduction, K-Means Clustering for customer segmentation, and Random Forest for customer classification, using primary data from the Padang City Water Company's Customer Meter Reading Report with an initial amount of 371 data. The results of the study show that the clustering process successfully formed three customer segments, namely premium customers with high consumption bills, regular customers with moderate and stable consumption, and new customers with low consumption rates. The evaluation of the Random Forest model's performance resulted in an accuracy rate of 68.85% on the training data and 67.69% on the testing data, with an average precision value above 0.84 and an average F1-score value of around 0.68. The consistency of performance between the training data and the testing data shows that the model has fairly good generalization capabilities and does not experience overfitting.

Implementation of K-Means Algorithm and C4.5 Classification in the Analysis of Determinants of Student Timely Graduation

Rahma Yanti — 2026-03-30

This study was motivated by the importance of timely graduation as a key parameter affecting program accreditation. The timely graduation rate reflects the effectiveness of academic management and serves as an indicator of program quality. The purpose of this study was to apply the concept of data mining using the K-means and Decision Tree C4.5 methods to analyze the timely graduation of students in the Information Technology and Computer Education Study Program at UIN Bukittinggi. The research methods used are the K-Means and Decision Tree C4.5 methods. The K-Means algorithm is used to cluster student graduation data, which will then be processed in the next method. The Decision Tree C4.5 algorithm is used to classify student graduation data. The research data was sourced from the 2017 batch of the Information Technology and Computer Education Study Program at UIN Bukittinggi, with a total of 158 data points. The results of this study produced a model that was able to achieve an accuracy rate of 96% in the validation process. The accuracy results were relatively high, so the model produced can be used by the study program to improve academic quality. Based on the results of this study, it contributes as a basis for evaluating student academic performance, monitoring the risk of study delays, and supporting academic decision-making. In addition, this information contributes to maintaining and improving academic quality and supports the achievement and maintenance of the accreditation status of the PTIK UIN Sjech M. Djamil Djambek Bukittinggi Study Program.

Public Sentiment Analysis of Train Services Based on Twitter Opinions Using K-Menas and SVM Methods

Dina Selvia — 2026-03-30

The development of social media, particularly Twitter, has become a primary means for the public to express opinions, criticisms, and complaints regarding train services, ranging from delays, facility comfort, to ticket policies. The large number of opinions appearing in short, non-standard characters, and containing slang and emoticons makes manual analysis ineffective, resulting in service providers not optimally utilizing valuable information from the public. This study aims to analyze public opinion sentiment on Twitter regarding train services to systematically and structuredly determine public perceptions. The methods used in this study are K-Means Clustering and Support Vector Machine (SVM). K-Means is used to group public opinion based on similarities in language patterns and sentiments to obtain initial labels, while SVM is used to classify opinions into positive and negative sentiments more accurately. The research data comes from the Twitter platform and is obtained through a crawling technique. The maximum limit of tweets retrieved is set at 2005 tweets. The results show that the K-Means method is able to assist the initial labeling process of sentiment data, while the SVM algorithm can classify public opinion with an accuracy level of 99.02%. The combination of clustering and classification methods has proven effective in processing large-scale, unstructured opinion data. Based on the research results, it can be concluded that the sentiment analysis approach using K-Means and Support Vector Machines can provide an objective picture of public perception of train service quality. The results of this analysis are expected to be used by service providers as evaluation material and a basis for decision-making to improve service quality to the public

Comparison of Decision Tree and Random Forest Methods in Predicting Oil Palm Productivity After Replanting

Sukardi — 2026-03-30

Oil palm is a strategic commodity in Indonesia that can be affected by various factors such as plant age, soil conditions, rainfall, and maintenance variations between farmers. Over time, oil palm productivity decreases, so it is necessary to predict the productivity of oil palm rejuvenation. Based on this, the purpose of this study is to apply and compare the Decision Tree and Random Forest algorithms to predict the level of oil palm productivity after rejuvenation. The prediction process was carried out at the Koperasi Unit Desa (KUD) Tirta Kencana, Kuantan Singingi Regency. The Decision Tree algorithm is a supervised prediction model, meaning it requires a training dataset whose role replaces past human experience in making decisions. The Random Forest algorithm is also able to present several decision trees used in the prediction process. The dataset in this study amounted to 241 farmer data sourced from the KUD Tirta Kencana in Kuantan Singingi Regency. The comparative results of these two methods show that both the Decision Tree and Random Forest algorithms are capable of predicting precisely and accurately. The comparative results show that the random forest method outperforms the decision tree method with an accuracy of 99%. The contribution of this research provides knowledge with the application of data mining science by comparing the performance of the decision tree and random forest algorithms in the process of plant productivity management at KUD Tirta Kencana.

Keywords: Oil Palm Productivity, Data Mining, Decision Tree, Random Forest, Productivity Prediction

System Usability Scale (SUS)–Based Evaluation of the PQX Study Program Archival System

Rahmi Elviana — 2026-03-30

Digital transformation has encouraged higher education institutions to implement archival information systems in order to improve the efficiency and effectiveness of records management. The PQX Study Program plans to utilize a digital archival information system as the primary medium for archival services; however, technical constraints and indications of low system usability have been identified. This study aims to analyze the usability of the archival information system in the PQX Study Program using the System Usability Scale (SUS) method. The research adopts a software engineering approach employing the Waterfall System Development Life Cycle (SDLC) model, which consists of requirements analysis, system design, prototype implementation, and evaluation stages. System testing was conducted through black box testing to verify system functionality, while usability evaluation was carried out using the SUS based on user perceptions. The results of the black box testing indicate that the application’s functional aspects are valid and operate as expected. Meanwhile, the SUS evaluation involving 10 respondents produced an average score of 60.25, which is below the SUS benchmark score of 98 and falls into category D. These findings suggest that the levels of effectiveness, efficiency, and user satisfaction remain suboptimal, indicating the need for improvements in interface design, process flow, and system usability support.