REAL ESTATE VALUATION USING ARTIFICIAL NEURAL NETWORK (ANN)

Transcription

1 REAL ESTATE VALUATION USING ARTIFICIAL NEURAL NETWORK (ANN) Prof. A.K. Soni 1, Abdulkadir Abubakar Sadiq 2 1,2 Dept. of computer Science &Engr. Sharda University Greater Noida, UP,(India) ABSTRACT This paper aims to demonstrate the importance and possible value of housing predictive power which provides independent real estate market forecasts on home prices by using artificial neural network. A feed forward back propagation (FFBP) network model was used in this research. We estimate the median value of owner occupied homes in Boston suburbs given 13 neighborhood attributes. An estimator can be found by fitting the inputs and targets. This data set has 506samples. Housing inputs is a matrix. The housing targets is a matrix of median values of owner-occupied homes in $1000 s. The result in this paper concludes that feed forward back propagation which is one of the network models in artificial neural network appears to be a good indicator of the output data to target data network structure than maximizing predict. The FFBP network which yield result from the Output_network for all samples are found from the equation output = 0.89 * Target The regression value is (R = 0.976). That means the Output_network is matching to the target data set (Median value of owner-occupied homes in $1000 s), and the percent correctly predict in the simulation sample is 97%. Keywords: Artificialneural Network, Real Estate Valuation and Feed Forward Back Propagation I. INTRODUCTION Today real estate market has become very popular. Though the near future of real estate is still in question, investors have been hungry for a fast way to play the market or to hedge against their volatile portfolios. Futures contracts have been an extremely popular method of balancing a portfolio in other markets, and real estate is, with a little Knowledge, now in the same boat [1]. Futures contracts that trade at a centralized exchange allow market participants more financial leverage and flexibility and are guaranteed by the exchange so there is no risk of counterparty default. They are also in and of themselves leveraged investments, which allow investors way to benefit on movements in housing prices as well as provide them with the opportunity for a liquid short term real estate investment. These futures also allow investors a way to speculate on housing prices with much lower capital requirements [1]. An accurate prediction on the house price is important to prospective homeowners, developers, investors, appraisers, tax assessors and other real estate market participants, such as, mortgage lenders and insurers [2]. Traditional house price prediction is based on cost and sale price comparison lacking of an accepted standard and a certification process. Artificial Neural Network (ANN) is a neurobiological inspired paradigm that emulates the functioning of the brain based on the way that neurons work, because they are recognized as the cellular elements responsible for the brain information processing [3]. ANN models can detect patterns that relate input variables to their corresponding outputs in complex biological systems for prediction [4]. 99 P a g e

2 Methods for improving network performance include finding an optimum network architecture and appropriate number of training cycles, using different input combinations [5]. Therefore, the availability of a house price prediction model helps fill up an important information gap and improves the efficiency of the real estate market [6] II. PROBLEM STATEMENT Today real estate market has become very popular. Though the near future of real estate is still in question, investors have been hungry for a fast way to play the market or to hedge against their volatile portfolios. Therefore an accurate prediction on house prices is important to prospective home owners, developers, investors, appraisers, tax assessors and other real estate market participants, such as, mortgage lenders and insurers. III. OBJECTIVE Here our main objective is to use feed forward back propagation (FBBP) network model to estimate or in other word predict the values of the house samples given based on the geographical variables as mentioned in the paper below. IV. ARTIFICIAL NEURAL NETWORK Neural network is an artificial intelligence model originally designed to replicate the human brain s learning process. The model consists of three main layers: input data layer (example the property attributes), hidden layer(s) (commonly referred as black box ), and output layer [7]. Neural network is an interconnected network of artificial neurons with a rule to adjust the strength or weight of the connections between the units in response to externally supplied data Figure 2[8,9]. Each artificial neuron (or computational unit) has a set of input connections that receive signals from other computational units and a bias adjustment, a set of weights for input connection and bias adjustment, and transfer function that transforms the sum of the weighted inputs and bias to decide the value of the output from computational unit [10]. In supervised training, we present a pattern to the neural network, it makes a prediction, and we compare the predicted output to the desired output. Thus we have explicit information about the performance of the network. The major parameters used in supervised training have to do with how the error is computed and how big a step we take when adjusting the connection weights in the direction of the desired output [11]. Learning rate almost all neural network models have a learning rate parameter associated with them. In a typical supervised training case, a pattern is presented to the neural network; it makes an incorrect prediction, and the difference between the desired output and the actual output is used to adjust the weights [12] There are many applications where prediction can help in setting priorities. For example, the emergency room at a hospital can be a hectic place. To know who needs the most time critical help can enable a more successful operation. Basically, all organizations must establish priorities which govern the allocation of their resources projection of the future is what drove the creation of networks of prediction [13]. 100 P a g e

3 Figure 1. Typical Structure of An Artificial Neural Network V. THE FEED FORWARD BACK PROPAGATION (FFBP) NETWORK A feedforward neural network is an artificial neural network where connections between the units do not form a directed cycle. This is different from recurrent neural networks. The feedforward neural network was the first and simplest type of artificial neural network devised. In this network, the information moves in only one direction, forward, from the input nodes, through the hidden nodes (if any) and to the output nodes. There are no cycles or loops in the network. Figure 2. The Structure of a Feed Forward Neural Network VI. RESEARCH METHODLOGY The methodology used is feed forward back propagation (FFBP). The data set consist of 506 samples, 80% is used for training, 10% is used for testing and the remaining 10% is used forvalidation in a feed forward manner. Weights and biases were randomly initialized. The network was trained with up to 100 epochs. Weight is information used by neural network to solve a problem. The dataset was gotten from the UCI Machine Learning Repository in Irvine, CA: University ofcalifornia, Department of Information and Computer Science, and the StatLib library which is maintained at Carnegie Mellon University. (1) This data set has 506 samples. (2) "Housing Inputs" is a 13x506 matrix with these rows: (3) Housing targets is a matrix of median values of owner-occupied homes in $1000 s. Below are the variables of the dataset (1) Per capita crime rate per town 101 P a g e

4 (2) Proportion of residential land zoned for lots over 25,000 sq. ft. (3) Proportion of non-retail business acres per town (4) 1 if tract bounds Charles river, 0 otherwise (5) Nitric oxides concentration (parts per 10 million) (6) Average number of rooms per dwelling (7) Proportion of owner-occupied units built prior to 1940 (8) Weighted distances to five Boston employment centers (9) Index of accessibility to radial highways (10) Full-value property-tax rate per $10,000 (11) Pupil-teacher ratio by town (12) 1000 (Bk )^2 (13) Percent lower status of the population VII. IMPLEMENTATION The neural networks were created using the neural network toolbox from Matlab 7.9. In this research. Training network automatically stops when generalization stops improving, as indicated by an increase in the mean square error (MSE) of the validation samples. The following step are as follow to create a neural network (1) Network size (2) Training (3) Validation (4) Testing VIII. RESULT After the specifications of the mentioned implementation by the user, the feed forward back propagation results is generated using the neural network toolbox on matlab. The below figure shows the result generated from the model mentioned above. Figure 3. The Best Validation Performance (MSE) of FFBP is at 14 Epoch 102 P a g e

5 Figure 4. THE TRAINING STATe for FFBP Network Figure 5. The Regression Values Between the Actual Value and Target Values (FFBP) The above regression plots display the network outputs with respect to targets for training, validation, and test sets. For a perfect fit, the data should fall along a 45 degree line, where the network outputs are equal to the targets. For this problem, the fit is reasonably good for all data sets, with R values in each case of 0.93 or above. If even more accurate results were required, you could retrain the network by clicking Retrain in nftool. This will change the initial weights and biases of the network, and may produce an improved network after retraining. 103 P a g e

6 IX. ANALYSIS OF RESULT As can be observed from the results in figure 4.3, feed forward back propagation neural network structure gives good results because the validation performance value (MSE) mean square error isless. The regression graph is getting it from the scatter plot in Figure 11, the relation between theoutput and the target. Output_network for all samples are result from the equation Output =0.95*Target The regression value R = That means the output_network is matching to the target data set (Median value of owner-occupied homes in $1000 s), and the percent correctly predict in the simulation sample is 97%. X. FUTURE SCOPE A comparative study would be taking between the existing artificial neural network model used in this paper i.e. feed forward back propagation (FFBP) and cascade forward back propagation(cfbp) method in predicting real estate prices and see which is more accurate. Feed forward neural network has only one direction from the inputs nodes, data goes through the hidden nodes (if any) and to output nodes. On the other hand CFBP network have the ability to learn and train the patterns quickly, the network determines its own size and topology, and it retains the structure it has built if the training set changes. In CFBP structure each layer neuron relates to all previous layer neuron that gives CFBP network more training to adjust the weight and gives more accurate result depending on the output of the network that matches the target. XI. CONCLUSION An accurate prediction on house prices is important to prospective homeowners and everything that belongs to the real estate market. The FFBP neural networks are one of these tasks used to predict house prices. When we apply the FFBP neural networks methodology to predict housing price, the result is good when the prediction is based upon selected parameters, FFBP shows ability of the network to learn the patterns, artificial neural networks shows significant results on house prices prediction in the simulation samples above of 97%. XII. AKNOWLEDGEMENT Our thanks goes to all the faculties of Department of Computer Science, Sharda University who have contributed towards the development of this paper and the constant guidance and encouragement received from Prof A.K. Soni, my parents in person of Malami Abdulkadir, Zainab Abdulkadir, my Siblings Aisha, Samira, Rakiya and Fatima Abdulkadir and not forgetting (Dr) Rabiu Musa Kwankwaso who made this dream a reality. REFERENCES [1] An Introduction to Real Estate Futures, [2] C. A. Calhoun, Property Valuation Models and House Price Indexes for The Provinces of Thailand: 2000, Housing Finance International, Vol. 18, No. 3, 2003, pp [3] A. Araque, E. D. Martin, G. Perea, J. I. Arellano and W.Buno, Synaptically Released Acetylcholine Evokes Ca2+Elevations in Astrocytes in Hippocampal Slices, Journal of Neuroscience, Vol. 22, No. 7, 2002, pp P a g e

7 [4] P. D. Wasserman, Advanced Methods in Neural Computing, Van Nostrand Reinhold, New York, 1993, p.255. [5] R. S. Parmer, R. W. McClendon, G. Hoogenboom, P. D.Blankenship, R. J. Cole and J. W. Dorner, Estimation of Aflatoxin Contamination in Preharvest Peanuts Using Neural Networks, Transaction ASAE, Vol. 40, No. 3, 1997, pp [6] H. Demuth and M. Beale, Neural Network Toolbox for Matlab-Users Guide Version 4.1, The Mathworks Inc., Natrick, [7] J. Coakley and C. Brown, Artificial Neural Networks in Accounting and Finance: Modeling Issues, InternationalJournal of Intelligent Systems in Accounting, Finance andmanagement, Vol. 9, No. 2, 2000, pp [8] G. Papadourakis, Introduction to Neural Networks, Technological Educational Institute of Crete, Department of Applied Informatics and Multimedia, [9] P. B. Joseph, Data Mining with Neural Networks Solving Business Problems, McGraw-Hill Companies, Inc., [10] A. Dave and M. George, Artificial Neural Networks Technology, Kaman Sciences Corporation, Utica, 1992 [11] A. Dave and M. George, Artificial Neural Networks Technology, Kaman Sciences Corporation, Utica, 1992 [12] R. A. Chayjan and M. Esna-Ashari, Comparison between Artificial Neural Networks and Mathematical Models for Equilibrium Moisture Characteristics Estimation in Raisin, Agricultural Engineering International: The CIGR E-Journal, Vol. 12, [13] M. H. Beale, M. T. Hagan and H. B. Demuth, Neural Network Toolbox 7 User s Guide, Math Works, Inc. Natick, P a g e