In this project, we calculate the lattice thermal conductivity of ZrS2 and WS2 in its pristine and defective form. The whole idea of this project was to introduce a new method to calculate thermal conductivity in systems with more complex structures. In the traditional form, there was two different approaches toward calculating the lattice thermal conductivity. First was by using solution of the Boltzmann traort equation and the second, by using molecular dynamic simulations. for the first method, the high computational cost limits the application of this approach, and it is insufficient in large and complex systems. Although the second approaches have been a success in accurate prediction of the lattice thermal conductivities of semiconductor crystals, the application of this technique to systems with more complex structures such as defective or disordered ones is not realistic because of the rapid increase in computational cost with increasing system size and simulations must be much larger to reproduce a reasonable temperature gradient. Computational time can be reduced by using the empirical potential, but the accuracy is insufficient compared to that of DFT-based calculations. The application of machine learning techniques to solidstate physics has rapidly developed in recent years. These techniques are promising for thermal conductivity simulation. In our novel approach, we extract high-order interatomic force constants by hiphive package with powerful machine learning algorithms. After obtaining the forces constants, we calculate the lattice thermal conductivity by running classical MD simulation With GPUMD or the solution of the Boltzmann traort equation (BTE) by PHONO3PY. The materials we are investigating are ZrS2 and WS2 which are the members of Transition Metal Dichalcogenides (TMDCs) semiconductors. ZrS2 has a low lattice thermal conductivity which is a candidate for thermoelectric materials and in the monolayer form, is good for photocatalytic, photo-detectors and solar cells. Although, WS2 has a high lattice thermal conductivity in compare to other TMDCs and can be used for thermal management in nanoelectronic applications and a Promising material for light-emitting diodes and optical sensors. At first, we investigated the lattice thermal conductivity of ZrS2 in direct method by solving BTE as implemented in the PHONO3PY code. then by using extract FCs from HIPHIVE. The last method has less computational cost but the same accuracy. Then we constructed Force Constant Potential (FCP) for WS2. The extracted FCs were processed by using PHONO3PY and as the potential for MD simulations. We calculated the lattice thermal conductivity of WS2 using both BTE and MD performed with GPUMD which agree extremely well with previous calculations and also experimental measurements. Finally, we calculated lattice thermal conductivity in defective WS2. We find that in presence of the defect thermal conductivity reduced because the mean free path of phonons is reduced. This approach can be used for further studies. The presented results prove that the explained method is an efficient way of predicting the thermal conductivity in both complex and large systems, which paves the way for future studies.