This paper presents a multi-factor framework in order to investigate the effects on thermal (heating and cooling) energy consumption of the climate and building envelope for buildings in the design stage. This framework, called "PSACONN Mining", is based on the Particle Swarm Optimization (PSO), Ant Colony Optimization (ACO), and Artificial Neural Network (ANN), with data mining techniques. It was aimed to find envelope combinations that provide minimum thermal energy consumptions for each climate. Firstly, attribute reduction was applied for climatic parameters according to thermal energy consumptions in order to provide more consistent results by eliminating irrelevant attributes. Then, energy simulation models were generated by using a certain number of building envelope combinations, which consist of various building materials. Finally, an ANN model based on PSO and ACO (PSACONN) was used in order to determine the most suitable envelope combinations by utilizing the reduced climate attributes and related envelope combinations along with thermal energy consumptions obtained through the simulation. PSO and ACO algorithms were used in order to train the ANN structure more appropriately by generating more consistent weight and bias matrices. The framework was applied to a public building in the design phase. It was observed that feature reduction improved the correlation between climate characteristics and thermal energy consumption by values of 0.98 and 0.96, respectively. When the heating and cooling consumption values obtained with the proposed framework are compared with the simulation values; The accuracy of the estimates for the heating and cooling energy consumption is above 99% and 98%, respectively. It is obviously demonstrated that, for different climate regions, the use of the proposed methodology gives effective results in finding envelope combinations that provide minimum thermal energy consumption.