Monument Preservation: Utilizing a Hybrid CNN-Random Forest Framework

This paper presents a hybrid CNN technique to develop an efficient model that categorizes monument preservation techniques into seven different classes. The proposed hybrid model plays an important role in monument preservation and also has a great contribution to increasing tourism on unknown monument sites as it trained on large data sets that increase the accuracy of this model. The model is trained on 3633 images and achieved an overall accuracy of 80.89%, the mico average of 80.90%, the macro average of 81.21%, a weighted average of 81%, and the highest f1-score of 85.81%. The study was conducted in the Punjab area that are useful for the preservation team to find the problems related to the monument and also attract the tourist attention toward its preservation. It minimizes the cost and reduces the time to find the solution related to preservation. The introduced hybrid model deeply recognizes the distorted or affected part of monuments and alerts the preservation team in the early stage. The findings of this hybrid model reduce the errors and it helps to achieve a feasible accuracy. In the future, the other hybrid approaches can be used for better accuracy.