Please use this identifier to cite or link to this item: http://hdl.handle.net/10266/5434
Title: Optimized Decision Making on Real Estate Data Using Data Analytics
Authors: Kaur, Gursimran
Supervisor: Kaur, Harkiran
Keywords: ETL;On-Line Analytical Processing (OLAP);Multidimensional Cube;Classification;j48 Decision Tree;Support Vector Machine (SVM);Non-Dominated Sorting Genetic Algorithm - III (NSGA-III)
Issue Date: 30-Oct-2018
Abstract: Buying or selling a property is a financial as well as an emotional undertaking. In this advanced era, these processes can be addressed differently than earlier, with more accuracy and optimization into them. In traditional times, real estate domain has been slow to embrace the recently emerged techniques. So, it’s a high time to start. Technologies such as machine learning can bring tangible benefits to all the parties involved. These may involve sellers, renters, buyers and tenants as well as brokers and agents. In this work, to make a CUBE without OLAP is clearly to form SQL queries that thinks the result sets (i.e., needed) and that contains comparable data (i.e., would come to execution in view of fuzzy OLAP exercises). There are some noteworthy burdens with this approach in this case. As an issue of first significance, the execution would be unacceptable when the database is broad with various relations required. Despite the fact that, the tests were performed with real time estimations and the audit reaction time was progressed. At the other hand, the queries were executed against the STAR Schema design which is stacked with abundance data to restrain the amount of joins required. To do comparable queries against a consistent data source would break down execution. Regardless, since the OLAP gadgets are especially created for these kind of queries, they are clearly enhanced for short query response times. A bit of these progressions abuse the read-generally nature of OLAP models and can hardly be found in a by and large valuable source database engine. Second, the reporting would be obliged. A great favored instance of OLAP gadgets is that the customer perceives is multidimensional and the documenting is achieved to a great degree of versatility. OLAP is outstandingly versatile with both segments, and paying little mind to whether it isn't so ordinary, uncovering more than two estimations is totally possible. Adding to this the roll-up and roll-down assignments improves these kind of gadgets than databases concerning separation of data. Clearly, the necessity for multidimensional data examination for a modest relationship with an obliged database may not require all the expansive furthest reaches of OLAP gadgets, which consistently are excessive, regardless of the fact that there are open source choices for Business Intelligence game plans too. This work proposes a real-estate mining process that is performed with the aid of J48 and Support Vector Machine (SVM) classification technique. Here, input dataset is high dimensional real-estate data which is a great barrier for classification. Therefore, initially feature dimension reduction using KPIs have been applied to reduce features space without losing the accuracy of classification. Here, unitary method has been used for selecting basic features from primary (self-created) dataset and secondary (taken from Kaggle website) datasets. Once the feature reduction is performed, the classification is applied based on J48 Decision tree and Support Vector Machine (SVM) classifier. From there on, the achieved information is changed into an arrangement issue that states whether the property has been acquired or not. To prepare the order information, J48 and SVM has been executed. In spite of the fact that, these models perform altogether to order land acquiring, at the same time, experience the ill effects of the parameter tuning issue. This issue has been settled by considering the outstanding meta-heuristic improvement strategy i.e., NSGA-III. It iteratively upgrades the meta-J48 model to enhance the classification rate by thinking about change using mutation and crossover operations. The acquired arrangements are non-dominant in nature, consequently, proposed model can give better accuracy as well as different parameters simultaneously. Broad experiments have been performed. It has been discovered that the proposed method beats as far as Accuracy, True Positive Rate, True Negative Rate, Precision and F_Measure. Consequently, the proposed strategy is relevant for ongoing land clients.
URI: http://hdl.handle.net/10266/5434
Appears in Collections:Masters Theses@CSED

Files in This Item:
File Description SizeFormat 
801631007.pdf4.3 MBAdobe PDFThumbnail
View/Open


Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.