Muhammad Fairus, Ramadhani (2026) Implementation of Content-Based Filtering and HDBSCAN for Developer Recommendation in a Task Management Application (Case Study: PT Tunas Kreasi Digital). Undergraduate thesis, UPN VETERAN JAWA TIMUR.
|
Text (Cover)
21081010090.-cover.pdf Download (640kB) | Preview |
|
|
Text (Bab 1)
21081010090.-bab1.pdf Download (123kB) | Preview |
|
![[img]](https://repository.upnjatim.ac.id/style/images/fileicons/text.png) |
Text (Bab 2)
21081010090.-bab2.pdf Restricted to Repository staff only until 22 May 2028. Download (325kB) | Request a copy |
|
|
Text (Bab 3)
21081010090.-bab3.pdf Restricted to Repository staff only until 22 May 2028. Download (4MB) | Request a copy |
|
|
Text (Bab 4)
21081010090.-bab4.pdf Restricted to Repository staff only until 22 May 2028. Download (1MB) | Request a copy |
|
|
Text (Bab 5)
21081010090.-bab5.pdf Download (105kB) | Preview |
|
|
Text (Daftar pustaka)
21081010090.-daftarpustaka.pdf Download (145kB) | Preview |
|
|
Text (Lampiran)
21081010090.-lampiran.pdf Restricted to Repository staff only Download (809kB) | Request a copy |
Abstract
The process of assigning developers in software development is generally still performed manually, making it prone to subjective bias and inefficiency. This study implements a developer recommendation system based on Content-Based Filtering (CBF) combined with the HDBSCAN (Hierarchical Density-Based Spatial Clustering of Applications with Noise) algorithm in a web-based task management application at PT. Tunas Kreasi Digital. The data used were obtained from GitHub commit logs and the company’s daily logs for the 2024–2025 period.The system works by clustering historical tasks using HDBSCAN based on the similarity of task description content, then applying Content-Based Filtering to recommend developers who previously handled tasks within clusters relevant to new tasks. The optimal HDBSCAN configuration was achieved with min_cluster_size = 2 and min_samples = 2, resulting in a Silhouette Score of 0.559 and the lowest noise percentage of 18.71%. Under this configuration, the system achieved a Hit Ratio of 0.88, Recall of 0.80, and Mean Reciprocal Rank (MRR) of 0.563.Testing on cosine similarity thresholds showed that using the entire historical dataset without filtering (4,505 records) produced the highest Hit Ratio and Recall values. Compared to the pure CBF method, the hybrid CBF+HDBSCAN approach outperformed almost all evaluation metrics, demonstrating that clustering developer expertise patterns can significantly improve recommendation accuracy. This system is considered feasible to implement as a decision-support tool for developer assignment in software projects.
| Item Type: | Thesis (Undergraduate) | ||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Contributors: |
|
||||||||||||
| Subjects: | Q Science > QA Mathematics > QA75 Electronic computers. Computer science | ||||||||||||
| Divisions: | Faculty of Computer Science > Departemen of Informatics | ||||||||||||
| Depositing User: | Muhammad Fairus Ramadhani | ||||||||||||
| Date Deposited: | 22 May 2026 07:11 | ||||||||||||
| Last Modified: | 22 May 2026 07:11 | ||||||||||||
| URI: | https://repository.upnjatim.ac.id/id/eprint/52128 |
Actions (login required)
 |
View Item |