Thermosyphon is an effective heat transfer device which is widely used all over the world for its ease of use, feasible with different environmental challenges. In this research article, the experimentation with different boiling point of working fluids water and R134a has been used in modified thermosyphon. The modified thermosyphon comprises of the cone frustum attached between the condenser and adiabatic section to hold up for high heat inputs. It is noted from the experiment that the working fluids have unique heat transfer capability with regard to thermal properties for applied heat input. The thermal resistance for different fill ratios with water and R134a in thermosyphon was experimented and the optimality in the fill ratio is identified. Due to the modification in the condenser, R134a condenser performs better for both low and high heat inputs but water works well only in high heat inputs. To predict and compare the temperature outputs at the evaporator, adiabatic and condenser sections of the thermosyphon, machine learning algorithm and optimization technique has been deployed and the output is measures for its accuracy, false positive, predictive positive value and effective performance. It is noted both from the experimental and algorithmic approach that the experiment produces less false positive rate which is ≤ 2% and true positive rate which is ≥ 98%, accuracy of the outputs which are ≥ 98%. The optimized outcome also stabilizes the experimental setup strongly and generates an effective performance rate which is ≥ 95%.