Multitype Inhomogeneous Poisson Point Process Dengan Sparse Group Lasso: Studi Kasus Kriminalitas Di Kennedy

Putra, Rendi Andria Gita (2026) Multitype Inhomogeneous Poisson Point Process Dengan Sparse Group Lasso: Studi Kasus Kriminalitas Di Kennedy. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5003221018-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (2MB) | Request a copy

Abstract

Kriminalitas di wilayah perkotaan seperti Distrik Kennedy, Bogotá, memiliki karakteristik spasial yang kompleks dan tidak menyebar secara acak. Dua jenis kejahatan dominants, yaitu pencurian kendaraan dan perampokan, diduga dipengaruhi oleh beberapa faktor lingkungan, dan masing-masing mungkin memiliki karakteristik yang berbeda. Tantangan utama dalam memodelkan fenomena ini adalah dimensi data yang tinggi serta adanya multikolinearitas antar kovariat lingkungan. Penelitian ini bertujuan untuk memodelkan pola spasial kedua jenis kejahatan tersebut menggunakan Multitype Inhomogeneous Poisson Point Process (IPPP) dengan penalti Sparse Group Lasso (SGL). Metode ini dipilih karena mampu melakukan seleksi variabel secara simultan pada tingkat grup dan tingkat individu. Pengelompokan dilakukan berdasarkan Crime Pattern Theory (CPT) sehingga ada empat grup untuk masing-masing jenis kejahatan, yaitu Crime Attractors, Crime Generators, Paths, dan Edges. Data yang digunakan meliputi titik kejadian perampokan dan pencurian kendaraan di Kennedy tahun 2012-2018 dan delapan kovariat spasial. Kecenderungan parameter penalti optimal mengindikasikan perlunya regularisasi yang ketat untuk menyeleksi fitur secara simultan pada tingkat kelompok maupun individu kovariat, sehingga menghasilkan model yang parsimoni. Model mengeliminasi dimensi Edges dan Paths pada kasus pencurian kendaraan, namun mempertahankan seluruh kovariat pada kasus perampokan. Kovariat jarak ke lokasi dengan riwayat korban luka teridentifikasi sebagai prediktor risiko terkuat untuk kedua jenis kejahatan, sementara fasilitas transportasi publik secara spesifik meningkatkan risiko perampokan namun tidak berpengaruh pada pencurian kendaraan. Hasil ini menunjukkan bahwa meskipun berbagi ruang yang sama, kedua kejahatan memiliki mekanisme pembangkit yang berbeda.
====================================================================================================================================
Crime in urban areas, such as Kennedy District in Bogotá, exhibits complex spatial characteristics and is not randomly distributed. Two dominant types of crime, vehicle theft and robbery, are hypothesized to be influenced by various environmental factors, potentially displaying distinct characteristics. The primary challenge in modeling this phenomenon lies in the high-dimensionality of the data and the presence of multicollinearity among environmental covariates. This study aims to model the spatial patterns of these two crime types using a Multitype Inhomogeneous Poisson Point Process (IPPP) with a Sparse Group Lasso (SGL) penalty. This method was selected for its capability to perform simultaneous variable selection at both the group and individual levels. The grouping was conducted based on Crime Pattern Theory (CPT), resulting in four groups for each crime type: Crime Attractors, Crime Generators, Paths, and Edges. The data utilized includes occurrence points of robbery and vehicle theft in Kennedy from 2012 to 2018 and eight spatial covariates. The optimal penalty parameters indicate a necessity for strict regularization to achieve simultaneous group-wise and individual feature selection, yielding a parsimonious model. The model eliminated the Edges and Paths dimensions for vehicle theft, whereas it retained all covariates for robbery. The distance to locations with a history of injuries was identified as the strongest risk predictor for both crime types, while public transportation facilities specifically increased the risk of robbery but had no significant effect on vehicle theft. These findings confirm that despite sharing the same spatial domain, the two crimes possess distinct underlying generating mechanisms.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Kriminalitas, Multitype Point Process, Sparse Group Lasso, Crime, Multitype Point Process, Sparse Group Lasso
Subjects:	H Social Sciences > HA Statistics > HA30.6 Spatial analysis Q Science > QA Mathematics > QA278 Cluster Analysis. Multivariate analysis. Correspondence analysis (Statistics)
Divisions:	Faculty of Science and Data Analytics (SCIENTICS) > Statistics > 49201-(S1) Undergraduate Thesis
Depositing User:	Rendi Andria Gita Putra
Date Deposited:	31 Jan 2026 04:14
Last Modified:	31 Jan 2026 04:14
URI:	http://repository.its.ac.id/id/eprint/131361

Actions (login required)

View Item