Buildings exchange heat with different environmental elements: the sun, the outside air, the sky, and outside surfaces (including the walls of environmental buildings and the ground). To correctly account for building energy performance, radiative cooling potential, and other technical considerations, it is essential to evaluate sky temperature. It is an important parameter for the weather files used by energy building simulation software for calculating the longwave radiation heat exchange between exterior surfaces and the sky. In the literature, there are several models to estimate sky temperature. However, these models have not been completely satisfactory for the hot and humid climate in which the sky temperature remains overestimated. The purpose of this paper is to provide a comprehensive analysis of the sky temperature measurement conducted, for the first time, in Djibouti, with a pyrgeometer, a tool designed to measure longwave radiation as a component of thermal radiation, and an artificial neural network (ANN) model for improved sky temperature forecasting. A systematic comparison of known correlations for sky temperature estimation under various climatic conditions revealed their limited accuracy in the region, as indicated by low R2 values and root mean square errors (RMSEs). To address these limitations, an ANN model was trained, validated, and tested on the collected data to capture complex patterns and relationships in the data. The ANN model demonstrated superior performance over existing empirical correlations, providing more accurate and reliable sky temperature predictions for Djibouti’s hot and humid climate. This study showcases the effectiveness of an integrated approach using pyrgeometer-based sky temperature measurements and advanced machine learning techniques ANNs for sky temperature forecasting in Djibouti to overcome the limitations of existing correlations and improve the accuracy of sky temperature predictions, particularly in hot and humid climates.