Analisis Explainable AI untuk Klasifikasi Crop dan Non-Crop pada Data Satelit Earth Observation

Bagaskara, Andymas Narendra (2023) Analisis Explainable AI untuk Klasifikasi Crop dan Non-Crop pada Data Satelit Earth Observation. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 05111940000192-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only until 1 October 2025. Download (12MB) | Request a copy

Abstract

Perubahan iklim yang terus meningkat dapat menyebabkan terancamnya ketahanan pangan dalam skala dunia. Akibat dari kekeringan yang terjadi pada negara berkembang meningkat sebanyak 80%, akibatnya pertanian menanggung beban ekonomi sebanyak 26% akibat dari bencana perubahan iklim tersebut. Sebanyak 24% emisi gas rumah kaca yang disebabkan seiring penggunaan lahan oleh pertanian, dapat berdampak pada perubahan iklim sehingga dapat berdampak pada pertumbuhan tanaman. Oleh karena itu dihadirkan dataset CropHarvest untuk mendapatkan informasi mengenai lokasi lahan pertanian serta menanggulangi permasalahan tersebut. Namun pada penelitian sebelumnya metode yang digunakan belum optimal. Pada penelitian ini diusulkan metode Explainable AI dengan fokus pada Shapley Additive Explanations (SHAP) untuk mengetahui fitur-fitur yang berpengaruh terhadap kinerja metode yang diusulkan terhadap metode yang lainnya, yaitu Random Forest, Support Vector Machine, dan Deep Learning Random. Hasilnya menunjukkan bahwa metode Explainable AI dengan pendekatan SHAP dapat mengidentifikasi fitur-fitur yang berpengaruh terhadap klasifikasi jenis tanaman. Berdasarkan hasil keseluruhan evaluasi, model Random Forest menunjukkan performa terbaik dengan nilai akurasi sebesar 0,9473, nilai AUC-ROC sebesar 0,9563, serta keseimbangan antara nilai F1 sebesar 0,9078, nilai Precision sebesar 0,9041, dan nilai Recall sebesar 1,0.
===============================================================================================================================
Increasing climate change can threathen food security on a global scale. As droughts in developing countries increase by 80%, agriculture bears 26% of the economic burden of these climate change disasters. As much as 24% of greenhouse gas emissions are caused by agricultural land use, which can have an impact on climate change so that it can have an impact on crop growth. Therefore, the CropHarvest dataset is presented to obtain information about the location of agricultural land and overcome these problems. However, in previous studies the method used was not optimal. In this research, the Explainable AI method is proposed with a focus on SHAP to find out the features that affect the performance of the proposed method against other methods, namely Random Forest, Support Vector Machine, and Deep Learning Random. The results show that the Explainable AI method with SHAP approach can identify features that affect the classification of plant species. Based on the overall evaluation results, the Random Forest model shows the best performance with an accuracy value of 0,9473, an AUC-ROC value of 0,9563, and a balance between F1 value of 0,9078, Precision value of 0,9041, and Recall value of 1,0.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	CropHarvest, Deep Learning, Explainable AI, Klasifikasi, Machine Learning, Classification, CropHarvest, Deep Learning, Explainable AI, Machine Learning
Subjects:	Q Science > Q Science (General) > Q325.5 Machine learning. R Medicine > R Medicine (General) > R858 Deep Learning T Technology > T Technology (General) > T57.5 Data Processing
Divisions:	Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Informatics Engineering > 55201-(S1) Undergraduate Thesis
Depositing User:	Andymas Narendra Bagaskara
Date Deposited:	05 Aug 2023 13:47
Last Modified:	05 Aug 2023 13:47
URI:	http://repository.its.ac.id/id/eprint/101296

Actions (login required)

View Item