Article Per Year (5 Year)
p-Index From 2021 - 2026
0.408
P-Index
This Author published in this journals
All Journal Enrichment: Journal of Multidisciplinary Research and Development
Ade Kania Ningsih
Universitas Jenderal Achmad Yani, Indonesia
Humanities Economics, Econometrics & Finance Engineering Health Professions Social Sciences

Published : 2 Documents Claim Missing Document
Claim Missing Document
Check
Articles

Found 2 Documents
Search

 1 
Consumer segmentation using K-Medians algorithm on transaction data based on LRFMP (length, recency, frequency, monetary, periodecity) Akbar Dena Maulana; Ade Kania Ningsih; Gunawan Abdillah
Enrichment: Journal of Multidisciplinary Research and Development Vol. 1 No. 8 (2023): Enrichment: Journal of Multidisciplinary Research and Development
Publisher : International Journal Labs

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.55324/enrichment.v1i8.70

Abstract

Consumer loyalty plays a crucial role for companies, particularly under intense competition among firms, and successfully retaining loyal customers is decisive for sustained profitability. For this reason, customer-loyalty analysis is needed to identify each customer's level of engagement with the company. Within this analysis, consumer segmentation is an essential step for grouping customers with similar characteristics so that the marketing-management process can be targeted more effectively. This study aims to segment e-commerce customers according to their behavioural loyalty and to characterise each resulting segment as a basis for differentiated marketing strategies. The segmentation employs the LRFMP model (Length, Recency, Frequency, Monetary, Periodicity), which represents customer purchasing patterns through relationship length, the recency of the last transaction, transaction frequency, total monetary value, and purchase regularity. Clustering is performed with the K-Medians algorithm, which uses coordinate-wise medians and Manhattan distance and is therefore robust to the outliers and skewness that are common in transaction data. The dataset comprises the purchase-transaction history of an e-commerce platform spanning 373 days, from which 4,712 unique customers were obtained after preprocessing. Applying LRFMP analysis with K-Medians produced four clusters, containing 1,183, 1,221, 1,206, and 1,102 customers, respectively. Interpretation of the LRFMP profiles indicates that 25.1% and 25.6% of customers (Clusters 1 and 3, jointly 50.7%) show high loyalty potential, 23.4% show medium potential, and 25.9% show low loyalty potential. The four-cluster solution attained an average silhouette coefficient of 0.608, indicating reasonably well-separated clusters.
Classification of Sentiment Towards BPJS Services Using the C50 Algorithm Amellia Fahezha Cahyaningrum; Yulison Herry Chrisnanto; Ade Kania Ningsih
Enrichment: Journal of Multidisciplinary Research and Development Vol. 1 No. 8 (2023): Enrichment: Journal of Multidisciplinary Research and Development
Publisher : International Journal Labs

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.55324/enrichment.v1i8.71

Abstract

Social media provides a timely source of public feedback on services delivered by the Social Security Administering Body for Health (BPJS Health), a State-Owned Enterprise responsible for Indonesia's public health insurance program. This study aimed to evaluate the ability of the C5.0 algorithm to classify positive and negative sentiment toward BPJS services in Twitter data. Applied quantitative research with an experimental text-classification design was conducted using a secondary dataset obtained from Kaggle. The implemented database displayed 3,060 documents. Data were processed through cleaning, case folding, tokenization, filtering, stemming, and TF-IDF weighting, followed by C5.0 classification and confusion-matrix evaluation using an 80:20 split. The reported test matrix comprised 621 cases: 6 true positives, 579 true negatives, 4 false positives, and 32 false negatives. These values produced 94.2% accuracy, 60.0% precision, and 15.8% recall. Although the aggregate accuracy was high, the low recall shows that the model detected only a small proportion of the positive class and was strongly influenced by the majority class. Therefore, the current model demonstrates the technical feasibility of applying C5.0 to BPJS-related tweets but cannot yet be considered balanced or fully reliable for service evaluation. Future optimization should address class imbalance, verify dataset labeling, and report complementary metrics before the results are used to support BPJS service-improvement decisions.
Co-Authors Akbar Dena Maulana Amellia Fahezha Cahyaningrum Gunawan Abdillah Yulison Herry Chrisnanto