発展途上国における小規模農家の金融排除の原因である信用情報欠落の解消のために、衛星データと地上データの統合による、新規アプリケーションとしての社会実装が可能なデータ駆動型の農地評価及び信用評価の手法を構築した。具体的に、（１）衛星データとモバイルデータを統合した借用履歴に代わる小規模農家の信用評価モデルを機械学習により構築し、その妥当性を確認した。（２）衛星データを駆使した客観的な農作物収量予測及び小規模農家からの収集データ・情報の信頼性検証をはじめとした収穫量予測の評価手法を確立した。（３）信用評価者の要求事項に基づいた社会実装可能なアプリケーションとして信用評価モデルの運用方法を提案した。
This research targets establishing an alternative credit scoring which can be used to support lending decisions to extend credit services to small-scale farmers in a developing country. Suitable data, credit score analysis models, and scoring results were evaluated. Appropriate credit scoring factors were designed, factoring in the socio-economic context of the farmers concerned which were established through field work observation and interviews. Data relating to farm activities and farmer characteristics, collected by a mobile application, was applied to the development of algorithms to predict the designed credit scoring factors. Additional sources of data such as open geospatial data sources were also be considered to improve model performance as needed. Geospatial analysis and machine learning were the main analysis methods used. The former was used to analyze location data. Meanwhile, various machine learning algorithms were applied and compared to identify the most fitting.

研究種目 : 基盤研究 (B) (一般)
研究期間 : 2019～2021
課題番号 : 19H04100
研究分野 : システムズエンジニアリング