This study investigates the situation of households’ food security in Pakistan. Food security is a comprehensive concept surrounding the nature, security of the food supply, quality, food access problems, and proper food utilization. The world has been facing the contradiction of widespread food insecurity and undernutrition. Present studies indicated that Pakistan country is a low-income developing country with an income per capita. Pakistan is one of the lowest in the world, but it, in general, has the economic capability to import the required food. However, in Pakistan, most areas are still food insecure, mostly belonging to Sindh and Baluchistan provinces. This study observes the main features of determining Pakistan food security, particularly household income, household economic evaluation, employment status, household expenditure, section, region, head age, head gender, agriculture status, livestock status, etc. Are studied indicators to measure the household food security status, whether it has food secure or insecure? And want to look at what conclusions can practically be drawn out of analysis when conducted within a conceptual framework. In this study average daily kilocalories per capita consumed index is used to measure the household’s food security level. Ordinal logistic regression and multiple linear regression models are used for analysis. For ordinal logistic regression model divided the Pakistan households into four categories based on the food security index that is daily kilocalories per capita. The research of this study shows that the primarily peoples living in Baluchistan, Sindh lies in food insecurity. Some households of KPK province lie in the food insecurity category. For conducting this study (PSLM), 2018-2019 survey data is used for analysis. Classical ordinal logistic and multiple linear regression models and machine learning, which includes ordinal logistic regression and multiple linear regression models, are used to analyze household food security in Pakistan. The model is finalized for best prediction based on the minimum Standard Error of the coefficient.