Mesfin Fikre (PhD)Tesfahiwot Tefera2026-03-062026-03-062022-11-19https://etd.aau.edu.et/handle/123456789/7879Humanitarian aid organizations and governments are faced with the challenge of locating and identifying the poorest individuals for efficient and effective intervention. The old ways of poverty assessment through survey is expensive and have limitations. Surveys are conducted once in many years, are prone to biases by local surveyors’ and supervisors, respondents give biased and wrong responses specifically if they know data is collected for aid purpose. As census is unlikely, sample based data collection is not as accurate as what one can do with latest targeting approaches. Currently researchers are exploring the use of night time satellite imagery and mobile phone data analysis for targeting purpose. These approaches are also not without limitations. In developing countries the rural poor are not connected to electric grid and hence less feasible to be used for targeting. In this paper we explore the use of bank transaction activities of individuals in aid prone counties in Ethiopia. Based on IOM’s 2021 humanitarian aid seekers, 25 counties in Amhara Region of Ethiopia are selected. Over 23 million daily transaction records with an amount of over 18bln birr for 2020 and 2021 is acquired. The result is quite interesting. By analyzing how bank transaction datasets like deposits, withdrawals, and transfers etc. trends over time, it is possible to prioritize aid seeking places accurately and objectively. Such trends or seasonality can be used to derive other measurement indexes, to explore how these kinds of datasets can be used to improve humanitarian beneficiary targeting. Finally, adding bank transaction data analytics indexes into targeting aid beneficiaries’ models can improve predictions. The knowledge gained from this study could provide valuable insights into strategies for utilizing banking datasets to increase efficiency and effectiveness of targeting. Keywords: targeting, humanitarian, data analytics, forecasting, time series, machine learning, bank account transactionsenThesis