{"title":"Early Prediction System for Employee Attrition Company “XYZ” Using Support Vector Machine Algorithm","authors":"Wikke Alvina Medyanti, M. Faisal","doi":"10.24114/cess.v8i2.46494","DOIUrl":null,"url":null,"abstract":"Pergantian karyawan merupakan masalah yang signifikan bagi organisasi karena dapat berdampak negatif pada produktivitas dan kinerja. Dalam penelitian ini, dikembangkan sebuah model Support Vector Machine (SVM) untuk memprediksi pergantian karyawan berdasarkan dataset yang berisi berbagai atribut karyawan. Dataset tersebut telah melalui tahap pra-pemrosesan dengan melakukan pemetaan nilai-nilai kategorikal dan pengkodean one-hot. Fitur-fitur kemudian dibagi menjadi data latih dan data uji, serta dilakukan penskalaan menggunakan StandardScaler. Hasil penelitian menunjukkan bahwa model mencapai akurasi sebesar 88,4%. Presisi untuk karyawan yang tidak mengalami pergantian (non-attrition) tinggi, yaitu sebesar 89,3%, menunjukkan kemampuan model dalam mengidentifikasi dengan benar karyawan yang kemungkinan akan bertahan. Namun, presisi untuk karyawan yang mengalami pergantian (attrition) lebih rendah, sebesar 69,2%, mengindikasikan adanya ruang untuk perbaikan dalam mengidentifikasi karyawan yang berisiko mengalami pergantian. Recall untuk karyawan non-attrition mencapai 98,4%, menunjukkan kemampuan yang tinggi dalam mengklasifikasikan dengan benar, sedangkan recall untuk karyawan attrition sebesar 23,1%. Nilai F1-score juga mencerminkan kinerja yang lebih baik untuk karyawan non-attrition dibandingkan karyawan attrition. Secara keseluruhan, model SVM menunjukkan potensi dalam memprediksi pergantian karyawan, namun perlu dilakukan pengembangan lebih lanjut untuk meningkatkan identifikasi karyawan yang berisiko, sehingga memberikan wawasan berharga dalam pengambilan keputusan SDM dan strategi retensi.Employee attrition is a significant concern for organizations as it can have a negative impact on productivity and performance. In this study, a Support Vector Machine (SVM) model was developed to predict employee attrition based on a dataset containing various employee attributes. The dataset was preprocessed by mapping categorical values and performing one-hot encoding. The features were then split into training and testing sets, and scaled using the StandardScaler.The results showed that the model achieved an accuracy of 88.4%. The precision for non-attrition employees was high at 89.3%, indicating the model's ability to correctly identify employees who are likely to stay. However, the precision for attrition employees was lower at 69.2%, suggesting room for improvement in identifying employees at risk of attrition. The recall for non-attrition employees was 98.4%, indicating a high ability to correctly classify them, while the recall for attrition employees was 23.1%. The F1-score also reflected a better performance for non-attrition employees compared to attrition employees. Overall, the SVM model showed promise in predicting employee attrition, but further enhancements are needed to improve the identification of employees at risk, thus providing valuable insights for HR decision-making and retention strategies.","PeriodicalId":53361,"journal":{"name":"CESS Journal of Computer Engineering System and Science","volume":null,"pages":null},"PeriodicalIF":0.0000,"publicationDate":"2023-07-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"CESS Journal of Computer Engineering System and Science","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.24114/cess.v8i2.46494","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
引用次数: 0
Abstract
Pergantian karyawan merupakan masalah yang signifikan bagi organisasi karena dapat berdampak negatif pada produktivitas dan kinerja. Dalam penelitian ini, dikembangkan sebuah model Support Vector Machine (SVM) untuk memprediksi pergantian karyawan berdasarkan dataset yang berisi berbagai atribut karyawan. Dataset tersebut telah melalui tahap pra-pemrosesan dengan melakukan pemetaan nilai-nilai kategorikal dan pengkodean one-hot. Fitur-fitur kemudian dibagi menjadi data latih dan data uji, serta dilakukan penskalaan menggunakan StandardScaler. Hasil penelitian menunjukkan bahwa model mencapai akurasi sebesar 88,4%. Presisi untuk karyawan yang tidak mengalami pergantian (non-attrition) tinggi, yaitu sebesar 89,3%, menunjukkan kemampuan model dalam mengidentifikasi dengan benar karyawan yang kemungkinan akan bertahan. Namun, presisi untuk karyawan yang mengalami pergantian (attrition) lebih rendah, sebesar 69,2%, mengindikasikan adanya ruang untuk perbaikan dalam mengidentifikasi karyawan yang berisiko mengalami pergantian. Recall untuk karyawan non-attrition mencapai 98,4%, menunjukkan kemampuan yang tinggi dalam mengklasifikasikan dengan benar, sedangkan recall untuk karyawan attrition sebesar 23,1%. Nilai F1-score juga mencerminkan kinerja yang lebih baik untuk karyawan non-attrition dibandingkan karyawan attrition. Secara keseluruhan, model SVM menunjukkan potensi dalam memprediksi pergantian karyawan, namun perlu dilakukan pengembangan lebih lanjut untuk meningkatkan identifikasi karyawan yang berisiko, sehingga memberikan wawasan berharga dalam pengambilan keputusan SDM dan strategi retensi.Employee attrition is a significant concern for organizations as it can have a negative impact on productivity and performance. In this study, a Support Vector Machine (SVM) model was developed to predict employee attrition based on a dataset containing various employee attributes. The dataset was preprocessed by mapping categorical values and performing one-hot encoding. The features were then split into training and testing sets, and scaled using the StandardScaler.The results showed that the model achieved an accuracy of 88.4%. The precision for non-attrition employees was high at 89.3%, indicating the model's ability to correctly identify employees who are likely to stay. However, the precision for attrition employees was lower at 69.2%, suggesting room for improvement in identifying employees at risk of attrition. The recall for non-attrition employees was 98.4%, indicating a high ability to correctly classify them, while the recall for attrition employees was 23.1%. The F1-score also reflected a better performance for non-attrition employees compared to attrition employees. Overall, the SVM model showed promise in predicting employee attrition, but further enhancements are needed to improve the identification of employees at risk, thus providing valuable insights for HR decision-making and retention strategies.