Page 139 - The-5th-MCAIT2021-eProceeding

relationship strength is shown in the form of numerical values. The relationship between attributes and house
price attributes is shown in Table 2 below. The correlation value of this coefficient is between -0 and 1.

Table 2 Level of Attributes Relationship with Home Price Attribute

Attribute Correlation Matrix Level of Relationship
b_tingkat 0.34 Quite strong
luas_lot 0.16 Moderate
luas_lot_bgn 0.16 Moderate
thn_perjanjian -0 No correlation

In this study after the experiments were conducted, the Random Forest (RF) algorithm gave the lowest MSE
value of 0.00017549, followed by the Gradient Boosting Decision Tree (GBDT) algorithm and k-Nearest
Neighbors (k-NN) with a value of 0.00020321 and 0.00022385 respectively. The result of each algorithm is
shown in Table 3 below.

Table 3 Level of Attributes Relationship with Home Price Attribute

Prediction Model MSE
RF 0.00017549
GBDT 0.00020321
k-NN 0.00022385

5. Conclusion

This study employs machine learning techniques to develop a price prediction model for house in Selangor.
It uses a dataset of residential housing for a 5-year period from year 2015 until 2020. The regression model
performances of the models are compared with one another and the accuracy of the prediction in this study
assessed by checking the mean squared error score of the training model. The pre-processing method has been
made before the test and the data is divided into two parts which is the training set and the test set. Random
Forest (RF), Gradient Boosting Decision Tree (GBDT) and k-Nearest Neighbors (k-NN) model which is a
machine learning regression algorithm are used in this study. The correlation relationship between the
dependent and independent variables was determined through the Pearson Coefficient. In this relationship is
explained through a correlation matrix where the relationship is expressed in coefficient values. From the results
of the experiments conducted, the correlation relationship for the numerical type variable is moderate with the
house price dependent variable. Based on the experimental results, the Random Forest (RF) model is the best
predictive model and has high accuracy compared to the Gradient Boosting Decision Tree (GBDT) and k-
Nearest Neighbors (k-NN) models.

Acknowledgements

I would like to acknowledge with gratitude to my supervisor Dr Nor Samsiah for her guidance in the process
of this work. Special thanks to National Property Information Center (NAPIC) for the data set used in the
experiment.

E- Proceedings of The 5th International Multi-Conference on Artificial Intelligence Technology (MCAIT 2021) [126]
Artificial Intelligence in the 4th Industrial Revolution