An Idea Rejected!

As the world moves on to next level of technology, computers and machines need to become more interactive and intelligent. The problem domains have gone so vast that explicit description for each considerable case makes it highly difficult and error prone technique for programming. Therefore, the solution may not be achieved by Case Based Reasoning (CBR). The solution is given by concepts of artificial intelligence. Though intelligence is not very clearly defined, still programs may be given a status of somewhat intelligent or highly intelligent depending upon the behavioral outputs that they give according to the problem domains.

The major objective is to conceptualize the real estate cost estimations using the artificial intelligence concepts. Major concepts to be used fall under the category of Neural Networks, Fuzzy Logic and Genetic Programming. The real estate cost estimation possesses a challenge that there are various parameters that drive the markets which do not have specified metric for evaluation. But still the human agents are able to quote the prices by examining those parameters. They may not even know how they are able to give variable weightage to different parameters, it may be concluded that they do so by their experience. Thus the program needs certain level of intelligence to make accurate or nearly accurate estimations.

Such a system would essentially need Fuzzy Logic implementation since the parameters defining the domain are otherwise not determinable. Only relative relations may be determined using membership values.

Though for estimation to be automatically or intelligently determined, either or both neural networks and genetic algorithms can be used. Neural networks provide computing and learning capability through the proposed model of Multi-Layered Perceptron with Backpropogation whereas genetic algorithms provide the concept of Fitness Function to improvise upon the results. More techniques from one or both concepts may be used for implementation.

The scope of this paper is limited to conceptualization and development of algorithms for providing solution to the problem domain by means of supervised learning in neural network model and improvisation of result by fitness function concept of genetic algorithms which may be extended to complete implementation further.