Semi-Automated Land Parcel Plotting: a Machine Learning Approach Based on Geospatial Data Matching

The Indonesian cadastral quality improvement process faces huge challenges due to its huge land area and past mapping practices. The manual methods prove insufficient, risking land conflicts. This research proposed a solution based on machine learning to semi-automatically search the location for plotting land parcels using geospatial data matching. The model’s performance was evaluated using the Sample Data and the KKP Database. In the first test, the model achieved a precision value of 98.75% and a recall value of 88.27%. The second test, involving more complex data, generates a lower precision value of 91% and a recall value of 57%, due to issues like the homogeneous shape of candidate locations and overlapping rights. To enhance the model performance, textual matching was used. It resulted in the improvement of recall value to 91% and the consistent results of the precision value of 92%. This study is novel due to the limited use of the methodology in certain fields. It can accelerate the current process of cadastral quality improvement in Indonesia. For future research, it is recommended to test the model in different data conditions and combine it with automation on data acquisition.