Emergency food assistance enable people or communities respond to food insufficiency, in a situation where individuals or households do not have enough diet to meet their basic needs. It is caused by several factors such as socio-economic indicators, demographic characteristics, and environmental factors. According to the World Food Programme, the emergency food assistance status of households is categorized as normal, vulnerable, and most-vulnerable. This study developed a predictive model that identifies the emergency food assistance status of households using explainable artificial intelligence and a hybrid learning approach, characterizing the socio-economic status of populations and providing accurate estimates of the most vulnerable households, crucial for governments and humanitarian organizations to make informed decisions about humanitarian assistance. Explainable artificial intelligence was developed to unveil the black box and shed light on the decision-making process. In this study, data for the experiment were collected from North Gondar by FH Ethiopia. The data was preprocessed to obtain high-quality data suitable for analysis, to create a model that predicts the emergency food assistance status of households. This study followed a design science methodology to conduct six experiments with 11,649 instances and 18 attributes to construct the proposed model. The dataset was partitioned into 80% for training, 10% for validation, and 10% for testing. The experiments showed that the hybrid model performed better than Random Forest, Extra Tree, LSTM, MLP, and Extreme Gradient Boosting, achieving an accuracy of 94%. As a result, the hybrid model was selected to develop a prototype and determine the risk factors influencing the need for emergency food assistance. The key factors are socio economic and demographic characteristics. Key variables identified include household with a large number of livestock, Households with a chronically ill member, households that have experienced the destruction of their crops due to natural disasters, and household with fewer than two plow oxen are identified as the first, second, third, and fourth risk factors for households requiring emergency food assistance. The LIME showed that households with fewer livestock, chronically ill members, female heads, and orphan members were most vulnerable with a 97% prediction probability. The prototype hosted on PythonAnywhere received an 84% acceptance rate from experts. Keywords- Food insufficiency, food insecurity, explainable artificial intelligence, predicting household emergency food assistance, ensemble machine learning.