Relevance. The interest in finding sustainable alternatives to classic heating and cooling systems. Geothermal energy is used because of its lower environmental impact compared to the traditional sources of conventional energy. The research is caused by the need to examine the impact of various design parameters on the outlet air temperature and mean efficiency of the earth-air heat exchanger system. Aim. To examine the impact of various design parameters, such as pipe length, air velocity, pipe diameter, and inlet conditions, on the outlet air temperature and mean efficiency of the earth-air heat exchanger system. The numerical data was used to train the artificial intelligence-based an artificial neural networks algorithm which was applied to predict the air temperature for more expanded range of effecting parameters. Methods. Numerical analysis using ANSYS FLUENT to model a new geometrical configuration of the earth-air heat exchanger, using a spiral pipe design. The computations determined the outlet temperature and earth-air heat exchanger efficiency, were used to train an artificial neural networks algorithm for predicting outlet temperature across a broader range of parameters. Results and conclusions. The study examined the impact of several factors, such as pitch spacing, air velocity, pipe diameter, inlet air temperature, and length, on the efficacy of the system during the coldest months of 2023. We concluded that the essential depth for effective heat exchange in earth-air heat exchanger is 6 to 7 m for all specified depths; the temperature of the discharge air increases as the pipe diameter decreases; as the air velocity decreased, the temperature of the exit air increased; the spiral pipe increased pitch distance results in a drop in the earth-air heat exchanger  outlet air temperature; when air velocity rises, the mean efficiency η falls. The largest reduction was observed with a lower air velocity and a smaller pipe diameter. The largest drop in η between pipe diameters of 110 and 250 mm is around 21%; the application of the anartificial neural networks approach was justified as it could predict the air temperature with very good accuracy.