Publication Search

Customer attrition has become the focus of many businesses today – since the online market space has continued to proffer customers, various choices and alternatives to goods, services, and products for their monies. Businesses must seek to improve value, meet customers' teething demands/needs, enhance their strategies toward customer retention, and better monetize. The study compares the effects of data resampling schemes on predicting customer churn for both Random Forest (RF) and XGBoost ensembles. Data resampling schemes used include: (a) default mode, (b) random-under-sampling RUS, (c) synthetic minority oversampling technique (SMOTE), and (d) SMOTE-edited nearest neighbor (SMOTEEN). Both tree-based ensembles were constructed and trained to assess how well they performed with the chi-square feature selection mode. The result shows that RF achieved F1 0.9898, Accuracy 0.9973, Precision 0.9457, and Recall 0.9698 for the default, RUS, SMOTE, and SMOTEEN resampling, respectively. Xgboost outperformed Random Forest with F1 0.9945, Accuracy 0.9984, Precision 0.9616, and Recall 0.9890 for the default, RUS, SMOTE, and SMOTEEN, respectively. Studies support that the use of SMOTEEN resampling outperforms other schemes; while, it attributed XGBoost enhanced performance to hyper-parameter tuning of its decision trees. Retention strategies of recency-frequency-monetization were used and have been found to curb churn and improve monetization policies that will place business managers ahead of the curve of churning by customers.

The development of image processing technology today can create systems that are able to effectively recognize digital images, one of which is in the field of agriculture for plant disease identification. Citrus plants experience a decrease in productivity due to pathogen attacks on leaves such as Black Spot, Cancer, and CVDP so that disease identification is needed. The classification method that can be used to classify images is the K-Nearest Neighbor (K-NN) algorithm because it is simple and has high accuracy in image management. This study aims to implement and determine the performance of the K-NN algorithm in identifying citrus plant diseases based on leaf images. This research uses a dataset from the Kaggle website of 1,096 images. There are 12 research scenarios using the comparison between test data and training data as much as 4, namely (90% training data + 10% test data, 80% training data + 20% test data, 70% training data + 30% test data, 60% training data + 40% test data) and testing with 3 random state values (42, 32, 22). The results showed that the K-NN algorithm is very effective in identifying citrus plant diseases with the highest accuracy value in the 90% training data scenario and 10% test data with a value of K = 2 which is 98.5%.

Mitra Telur is an egg producer located in Pirakbulus, Sidumulyo, Godean District, Sleman Regency, Special Region of Yogyakarta. So far, Mitra Telur UMKM has not determined a travel route to distribute the eggs. The distribution carried out does not take into account the distance traveled to reach the shop points. This research uses two methods at once, namely saving matrix and nearest neighbor. Based on the calculation results, the initial route has a total distance of 114.9km with 4 delivery routes, while the final route has a total distance of 95.5km with 3 delivery routes. The initial route has a fixed cost of IDR 1,550,000 and a variable cost of IDR 402,150 with a total delivery cost of IDR 1,952,150, while the final route has a fixed cost of IDR 1,550,000 and a variable cost of IDR 334,250 with a total delivery cost of IDR 1,884,250. there was a reduction in distribution routes by 16.9% and a reduction in production costs by 3.5%.

Mitra Telur is an egg producer located in Pirakbulus, Sidumulyo, Godean District, Sleman Regency, Special Region of Yogyakarta. So far, Mitra Telur UMKM has not determined a travel route to distribute the eggs. The distribution carried out does not take into account the distance traveled to reach the shop points. This research uses two methods at once, namely saving matrix and nearest neighbor. Based on the calculation results, the initial route has a total distance of 114.9km with 4 delivery routes, while the final route has a total distance of 95.5km with 3 delivery routes. The initial route has a fixed cost of IDR 1,550,000 and a variable cost of IDR 402,150 with a total delivery cost of IDR 1,952,150, while the final route has a fixed cost of IDR 1,550,000 and a variable cost of IDR 334,250 with a total delivery cost of IDR 1,884,250. there was a reduction in distribution routes by 16.9% and a reduction in production costs by 3.5%.

This research relates to predictions of laptop sales in computer shops in Central Aceh, with a focus on laptop brands Acer, Asus, HP and Lenovo. Over the last three years, sales of these laptops have reached 1,629 units, with a monthly average of between 108 and 150 units. Business owners today prefer brands with the highest percentage of sales, but this can lead to dead stock problems. Therefore, the author proposes using data mining techniques, especially the K-Nearest Neighbor (K-NN) method, to make recommendations for the number of products to be purchased by business owners based on past sales data. The K-NN method requires complete, structured and continuous sales data. It is important to choose an appropriate K value, and other factors such as weather, seasons, promotions, and special events also affect laptop sales. K-NN models may need to be combined with other data to improve prediction accuracy. It is hoped that this research will provide academic benefits in expanding knowledge about the use of the K-NN method in sales prediction, as well as practical benefits for business owners in planning their sales strategies. The research conclusions highlight the importance of good data collection, choosing the right K value, and considering external factors in the laptop sales prediction process.      

Fraudsters increasingly exploit unauthorized credit card information for financial gain, targeting un-suspecting users, especially as financial institutions expand their services to semi-urban and rural areas. This, in turn, has continued to ripple across society, causing huge financial losses and lowering user trust implications for all cardholders. Thus, banks cum financial institutions are today poised to implement fraud detection schemes. Five algorithms were trained with and without the application of the Synthetic Minority Over-sampling Technique (SMOTE) to assess their performance. These algorithms included Random Forest (RF), K-Nearest Neighbors (KNN), Naïve Bayes (NB), Support Vector Machines (SVM), and Logistic Regression (LR). The methodology was implemented and tested through an API using Flask and Streamlit in Python. Before applying SMOTE, the RF classifier outperformed the others with an accuracy of 0.9802, while the accuracies for LR, KNN, NB, and SVM were 0.9219, 0.9435, 0.9508, and 0.9008, respectively. Conversely, after the application of SMOTE, RF achieved a prediction accuracy of 0.9919, whereas LR, KNN, NB, and SVM attained accuracies of 0.9805, 0.9210, 0.9125, and 0.8145, respectively. These results highlight the effectiveness of combining RF with SMOTE to enhance prediction accuracy in credit card fraud detection.

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 research is intended to design a waste bank information system through an action research approach. The object in this study is a group of waste banks under the auspices of the Guwosari Training Center, Bantul Regency. The new system built will strengthen ease of use (easy to use). Two things are proposed in this study, namely improvements to the waste bank manual information system that is already running as an initial diagnosis and the transition from the manual recording system to digital recording through computer devices (applications) with centralized data in one database.

Rice plays a vital role as the main food source for almost half of the global population, contributing more than 21% of the total calories humans need. Production predictions are important for determining import-export policies. This research proposes the XGBoost method to predict rice harvests globally using FAO and World Bank datasets. Feature analysis, removal of duplicate data, and parameter tuning were carried out to support the performance of the XGBoost method. The results showed excellent performance based on which reached 0.99. Evaluation of model performance using metrics such as MSE, and MAE measured by k-fold validation show that XGBoost has a high ability to predict crop yields accurately compared to other regression methods such as Random Forest (RF), Gradient Boost (GB), Bagging Regressor (BR) and K-Nearest Neighbor (KNN). Apart from that, an ablation study was also carried out by comparing the performance of each model with various features and state-of-the-art. The results prove the superiority of the proposed XGBoost method. Where results are consistent, and performance is better, this model can effectively support agricultural sustainability, especially rice production.