Publication Search

This paper compares various machine learning models in their ability to predict financial trends, with a focus on time-series analysis. We evaluate models such as linear regression, decision trees, support vector machines, and deep learning, measuring their performance based on accuracy, computational cost, and interpretability. Our results reveal that deep learning models offer superior accuracy but are less interpretable, while simpler models, though less accurate, provide better insight into the underlying data. This research provides guidelines for selecting suitable models based on specific financial applications.

The occurrence of scorch during the production of flexible polyurethane is a significant issue that negatively impacts foam products' resilience and generally jeopardizes their integrity. The likelihood of foam product failure can be decreased by optimizing production variables based on machine learning algorithms used to predict the occurrence of scorch. Investigating technology is required because prevention is the best approach to dealing with this problem. Hence, machine learning algorithms were trained to predict the occurrence of scorch using the thermodynamic profile of polyurethane foam, which is made up of recorded production variables. A variety of heuristics algorithms were trained and assessed for how well they performed, namely XGBoost, Decision trees, Random Forest, K-nearest neighbors, Naive Bayes, Support Vector Machines, and Logistic Regression. The XGboost ensemble was found to perform best. It outperformed others with an accuracy of 98.3% (i.e., 0.983), followed by logistic regression, decision tree, random forest, K-nearest neighbors, and naïve Bayes, yielding a training accuracy of 88.1%, 66.7%, 84.2%, 87.5%, and 67.5% respectively. The XGBoost was finally used, yielding 2-distinct cases of non(occurrence) of scorch. Ensemble demonstrates that it is quite capable and is an effective way to predict the occurrence of scorch.

This study aims to evaluate the application of a new methodology in investment decision-making, specifically using the regression tree approach on stock market indices. This approach is expected to enhance prediction accuracy and assist investors in making more informed investment decisions, especially in volatile and uncertain markets. Based on the literature review, regression trees offer advantages in identifying non-linear relationships between market variables that are often undetected by traditional models such as the Capital Asset Pricing Model (CAPM). Despite its advantages, the application of regression trees also faces challenges, such as overfitting issues and the need for large and complex data. This study concludes that regression trees can improve investment decision-making, but careful attention is required regarding model tuning and data quality.

As many as 7.3 million people worldwide die from heart disease. This indicates that heart disease is one of the diseases that cause the most deaths. As a preventive effort in handling heart disease, it is necessary to predict heart disease in patients. The classification process to predict heart disease is done using a decision tree. This decision tree is interesting because it is more flexible in providing the advantage of visualizing the advice so that the prediction can be observed. This study uses Heart Disease Prediction Dataset data with a total of 303 data. Then predictions are made using Decision tree so that the accuracy results are 83.60%, precision 89.28%, recall 78.12% and F1 score of 83.33%.

The use of the Decision Tree method in smartphone price classification is the focus of this study. By using the 10 most relevant features and data normalization to achieve scale consistency, the Decision Tree algorithm delivers an average accuracy of 81%. Although some false positives and false negatives occur, the model is able to classify smartphone prices well, especially in identifying low and high prices. These results provide important insights into the features that affect smartphone prices. While there is still room for improvement, this model provides a solid foundation for the smartphone industry to determine prices based on certain specifications. The importance of relevant feature selection and data normalization was revealed in this study. Despite the accuracy reaching 81%, improvements in the classification of medium and high price classes are still possible to reduce prediction errors. This method provides an important basis for the smartphone industry to set prices based on specifications, and data mining techniques such as Decision Tree can be improved to improve the accuracy of future price predictions.

Di era digital saat ini, kemajuan teknologi informasi mengalami pertumbuhan yang pesat. Salah satu perkembangan yang signifikan terjadi dalam bidang kecerdasan buatan (Artificial Intelligence), yang telah diterapkan luas di berbagai sektor, termasuk analisis data dan pengambilan keputusan. Para peneliti di bidang data mining telah menciptakan beragam algoritma klasifikasi yang meningkatkan proses klasifikasi dengan memanfaatkan atribut numerik dan nominal. Klasifikasi adalah suatu proses analisis data yang menghasilkan model untuk mewakili kelas-kelas dalam data tersebut, seperti yang terjadi pada Decision Tree yang digunakan untuk menganalisis klasifikasi dan pola prediksi data serta menggambarkan hubungan antara variabel atribut  dan variabel target  dalam bentuk struktur pohon. Python, sebagai bahasa pemrograman populer dalam pengembangan kecerdasan buatan, menyediakan pustaka dan framework yang mendukung implementasi algoritma Decision Tree. Dengan menerapkan algoritma Decision Tree untuk klasifikasi data evaluasi mobil menggunakan Python, kita dapat memanfaatkan kekuatan AI untuk memberikan solusi efektif dan efisien dalam pengambilan keputusan terkait evaluasi mobil. Menggunakan dataset Evaluation Car dari UC Irvine Machine Learning Repository, hasil penelitian menunjukkan akurasi sebesar 81%. Confusion Matrix dan laporan klasifikasi menunjukkan performa model yang baik dalam melakukan prediksi.

The goal of this data mining C4.5 implementation is to improve student performance in academic coursework in the computer science department at Teknik Fakultas and Pancasakti University in Tegal. Use a limited number of dimensions to assess the following: nyata, jaminan, keandalan, empatia, dan bukti nyata. It is difficult to determine which quality standard has to be raised because the aforementioned kelima aspek cannot be changed in an objective manner. Utilizing the algorithm C4.5 method, the authors consider reducing the sample size to the point where the keputusan is reduced. After manual perhitungan, pembuktian is also carried out using an application called RapidMiner.The analysis's conclusions show that the most important factor in determining the mahasiswa's tingkat kepuasan is the style of teaching.

Data mining is a technology that plays an important role in supporting data-driven decision making, especially in complex and dynamic higher education environments. In the context of education management, the ability to predict student graduation is an essential aspect because it can help institutions plan strategic steps, intervene earlier, and optimize academic resources. This study aims to apply the C4.5 decision tree algorithm to build a student graduation prediction model based on academic data. The research dataset includes key variables such as Grade Point Average (GPA), total Semester Credit Units (SKS) taken, and student attendance rates during lectures. The analysis was conducted using the C4.5 algorithm, which is known for its high level of interpretability, making the model results easy to understand by policy makers. The test results showed an accuracy of 84.6%, indicating that this method has the potential to support data-based academic management systems. These findings are expected to serve as a basis for educational institutions to improve the effectiveness of monitoring and evaluating the student learning process.

. In the shipbuilding industry, the repair process is a series of jobs that require a relatively short time. Delays in repairs can occur due to weak management and also caused by less than optimal empowerment of human resources. This study aims to identify the risks found in four divisions, namely the Production Division, Warehouse Division, Finance Division and Purchasing Division. The study found 31 risk events. Determining the value of each risk is carried out using the Failure Mode and Effect Analysis method. There were 13 risks that had the highest Risk Priority Number, namely the length of approval for requests for goods (RPN = 522.88), delays in payment processes by customers (RPN = 504.64), requests for additional work from the owner (RPN = 477.128), fluctuations in the number of manpower (RPN = 454.08), Changes in material use related to the availability of materials in the warehouse (RPN = 411.768), Changes in material calculations related to design (RPN = 389.017), Length of decision making by the owner (RPN = 388.36), Long material delivery process (RPN = 388.36), Insufficient stock material (RPN = 357,588), Writing the amount on the Request for Goods Bill is not detailed (RPN = 357.71), Lack of availability of stock material (RPN = 349,524), Making and submitting late payment requests (RPN = 316.8), Placement of materials that are less efficient (RPN = 296.8) Risk mitigation is carried out using the Fault Tree Analysis method to find the main cause / basic event of each risk. And the mitigation step that needs to be done is by making changes to the warehouse layout. If the layout design of the warehouse is changed to be more efficient it will speed up the material retrieval process which has an impact on the ship repair process time.

The Mediterranean region, in particular Algeria, is experiencing serious challenges due to the increased opportunities for forest fires. Since the mid-1970s, there has been a 50% reduction in rainfall over northwestern Algeria, making northern Algeria particularly vulnerable to the problem for many years. More than 37,000 hectares of sensitive forest are lost every year due to this extreme drought. The findings of this study, which assessed the hazard of forest fires from 2006 to 2019, agree with those of Bentchakal,Chibane (2022), who examined the problems caused by forest fires in the region. The aim of this investigation is to gain a better understanding of the problems caused by local forest fires and to use that expertise to provide insight for the authors and readers of this report. The report was written by presenting the findings of observations made using the Rapid Miner classification approach, which includes the categorization of areas affected by forest fires. Data is collected using a variety of algorithmic techniques, including Naive Bayes, KNN, and decision trees, which are used as tests of data to identify the most accurate results. The findings show that the Decision Tree technique has the best accuracy of 86.49% and provides a thorough explanation of the data.

Perguruan  tinggi di Indonesia saat ini sudah mencapai 2000 lebih perguruan tinggi.  Perguruan tinggi bisa tetap eksis salah satu faktornya adalah dengan adanya mahasiswa yang masuk ke perguruan tinggi tersebut. Dalam penerimaan mahasiswa baru setiap perguruan tinggi tentunya tidak sembarangan, mereka akan melakukan seleksi terhadap calon mahasiswa agar mendapatkan mahasiswa dengan kualitas yang mereka harapkan. Ada beberapa cara dalam melakukan seleksi calon mahasiswa, diantaranya ada yang berdasarkan nilai rapor, ada yang melalui tes tertulis, dan ada jalur mandiri. Tidak semua perguruan tinggi melakukan pengolahan data yang masuk secara maksimal. Biasanya mereka masih mengolah data secara manual atau dengan kata lain belum menggunakan machine learning. Masalah yang ada yaitu adanya perguruan tinggi yang belum secara maksimal dalam mengolah data penerimaan calon mahasiswa. Karena belum optimalnya data yang diolah maka ada perguruan tinggi yang belum dapat membuat prediksi dalam penerimaan calon mahasiswa untuk di tahun ajaran berikutnya. Salah satu untuk dapat membuat prediksi di masa mendatang adalah dengan menggunakan data mining.  Metode yang akan digunakan dalam data mining adalah klasifikasi Decision Tree dioptimasi dengan Particle Swarm Optimization. Hasil dari Decision Tree dengan optimasi Particle Swarm Optimization didapatkan akurasi sebesar 99,94%.

With the increasing development of technology the more variety of books circulating on the internet. As is the recommendation system on online book sites that provide books relevantly and as needed with one's preferences. One alternative is GoodReads, a social networking site that specializes in cataloging books and users can share reading book recommendations with each other by rating, reviewing, and commenting. As a large book recommendation site, it has a lot of data that can be processed by applying machine learning methods, but still not known as the most accurate model. By using the right model, we can provide more accurate recommendations. Therefore, this study will analyze the data obtained from the www.kaggle.com namely the goodreads-books dataset. This study proposed a data mining classification model to get the best model in recommending books on GoodReads. The algorithms used are Decision Tree, K-Nearest Neighbor, Naïve Bayes, Random Forest, and Support Vector Classifier, then for model evaluation using accuracy, precision, recall, f1-score, confusion matrix, AUC, and Mean Error Absolute. The test results of several classification algorithms found that Decision Tree has the highest accuracy among the methods presented by 99.95%, precision by 100%, recall by 96%, f1-score of 98% with MAE of 0.05 and AUC of 99.96%. This is proof that decision tree algorithms can be used as book recommendations based on book categories on GoodReads.

The shift in global paradigm and threat perspective has led to a wide variety of possible risks and uncertainties. This situation also occurs in the defense economy, so understanding the basic principles of risk and uncertainty is important, especially in a decision-making process. There are several elements and concepts that are usually used in all decision models. Almost all models, whether complex or simple, can be formulated using a standard structure and solved by using general evaluation procedures. For decisions involving a series of decisions and relating to various basic sequential conditions, the decision tree is an appropriate conceptual and schematic modeling tool. A decision tree is a schematic representation of a decision problem. A decision tree is a diagram made like a tree with branches and branches in a chronological order of events, with each having a choice and possibility of occurrence, as well as the results of each choice. The term decision tree is taken from the form of diagrams that have branches and twigs, just like a tree.

Infant health can be known one of them through the assessment of nutritional status. In general, Body Mass Index (BMI) has been used as a method for measuring the nutritional status of children. If there are two children who have same body weight and height, they may have different nutritional status. Whenever this occurs, the use of BMI for measuring the nutritional status shall be deemed less accurate. The anthropometry will be vital in measuring the nutritional statuss. The guidelines for determining the nutritional status Anthropometry parameters are selected and recommended which includes an assessment of the age, weight, body length or height. This research aims to build a model of C4.5 adaboost so it can recognize patterns and be able to classify the nutritional status of children into five classes: normal, fat, very fat, thin and very thin. The variables used in this classification is Gender, Age (Months), Weight (kg) Height (cm). C4.5 (decision tree) Method has a good performance in dealing with the classification of nutritional status but the C4.5 has a weakness in the class imbalance. Adaboost isone ofboosting methods that could reduce imbalances class by giving weight to the level of classification error which may alter the distribution of data. Experiments carried out by applying the adaboost method C4.5 to obtain optimal results and a good degree of accuracy. The experimental results obtained from C4.5 method show that accuracy is 89.53%, the error rate is 10.47%, while the results of C4.5 with adaboost show 90.23% accuracy and 9.77% error rate. It can be concluded in the classification of nutritional status of children with C4.5 and adaboost proven method to solve problems of class imbalance and improve the high accuracy and can reduce the level of classification error.

Academic data increases every year in line with the increase of students. Abundant data store is alsoan abundance of information. Data mining technology is a tool for extracting information on largedatabases and has been widely used in many domains. Predicting student performance (study evaluation) isan activity to determine a future state based on existing data. Data in the field of academic research hasbeen done with various methods and algorithms, but the use of algorithm SLIQ (Supervised Learning InQuest) has not been done.SLIQ is an algorithm developed by the IBM's Quest project team in 1996 for mining large datasets.SLIQ algorithm classify and predict the students performance, beginning with the data cleaning, conductedelection training and testing data. By calculating gini index of each attribute and then selecting thesmallest gini index data table is split according to the criteria until find the same class. From the results ofthe calculation process can produce a set of rules that can be used to predict student performance.From the experiment it can be concluded that the algorithm SLIQ with decision tree technique canbe used as an alternative in designing a system datamining applications. Tests conducted system showedthat the constructed model can be used to predict the performance of new students. The resulting accuracyof the model system in fact has a lower score than the accuracy of other applications that are used as acomparison of Tanagra. Advantages of the proposed system is in its design does not need complexcalculations in obtaining the gini index attributes.