I agree to submit this thesis/dissertation for research X. The research reported in this thesis, unless otherwise stated, is my original work. Limited studies have been conducted on types of behavior that drive property owners to sell their property.
Introduction
Motivation for the Study
This study will break down the various data sources related to the real estate industry in order to create predictive models, which will contribute to an improved understanding that will drive business strategy to determine property sales. . With this understanding, real estate agencies can improve and optimize their approach when generating leads by targeting specific consumers who have a high probability of selling their property.
Focus of the Study
Real estate companies need to utilize these marketing tactics to elicit a seller's response by motivating them to sell their property. This study will be conducted on seller behavior within the real estate industry geographically located in KwaZulu-Natal.
Significance of the Study
A large area of focus relates to the salesperson's life cycle in terms of "what drives" an individual to sell, e.g. This study contributes to the existing body of knowledge by enriching the current literature on what drives property owners to behave in certain ways when selling their property.
Problem Statement
In addition, the study aimed to contribute to the limited research conducted on residential property in KwaZulu-Natal by exploiting the abundance of consumer and property data.
Research Sub-Questions
Objectives
6 density, active businesses in the area, number of homes sold in the last 5 years, average real estate sales growth, and consumer life cycle stage. The business objective is to reduce sales effort, resource costs and create efficiencies in the sales process.
Research Hypotheses
7 H1c: The identified consumer population within the KwaZulu-Natal submarkets can be measured in terms of determining the propensity of a property owner to sell their property.
Methodology
Outline of Thesis
It explains the results in relation to the context of the literature review presented in the second chapter. This chapter discusses and presents the conclusions of the study as well as the limitations of the research and recommendations of the study.
Delimitation of the study
Limitations of the study
The second limitation in the study was missing data; especially address information from the Deeds data source.
Assumptions
Summary
Introduction
Consumer behaviour
- Definition of consumer behaviour
- Consumer choice and decision making
- Ethical analysis of targeting consumers’ needs and wants
- Five stage decision making process
- Problem recognition process
- Information search process
- Evaluation of alternatives
- Selling decision
- Post sale behavior
- Influences on consumer behavior
- Individual Influence
- Environmental influences
- The relationship between Consumer Behavior and the housing market
According to Nwanko et al. 2014), the involvement of the consumer in the purchase decision depends on the type of product and its relationship to the consumer, which dictates the type of information that the consumer has to process. Developed infrastructure in a residential area contributes to the rise or fall of house prices.
Integration of consumer behavior theories and the effects on real estate
Decision making process
Predictive models
- Implications of scorecard models within real estate include
Consumer Behaviour Models
Overview consumer behaviour models
Classifications of consumer models
Customer behaviour theoretical models
- Hierarchy of needs model
- Micro models
Marshall's Economic Model According to Marshall (1890), consumers will spend their income on the goods and services that will bring them the greatest satisfaction, depending on their tastes and the price of the goods. The point of view of this model is that economic factors should be included in a comprehensive description of buyer behavior, since economic factors have a significant effect on all markets. A reaction is a response to a suggestion, and the repetition of the reaction is influenced by an.
Cluster Analysis: Statistical Technique
Key clustering concepts
Supervised learning models: Supervised learning is based on training a data sample from a data source that has already been assigned the correct classification (Kalhori and Zeng, 2014). A supervised learning algorithm analyzes the training data sample and produces a derivative function that can be used to map new examples. Cluster Methods for Partitioning Clusters: According to Vijayarani and Jothi (2014), this method decomposes the data object sets into clusters where each pair of clustered objects is either distinct (hard clustering) or has some members in common (soft/fuzzy clustering).
Common clustering algorithms
Bottom-up Hierarchical Algorithm: According to Zhang et al (1996), this method starts clustering by placing each object in different clusters and then groups them into increasingly larger clusters until all objects are in the same cluster, based on similarities that are they are shared by clusters. DBScan Algorithm: According to Ester et al (1996), the density-based clustering algorithm segments the data based on the particle density of some threshold. According to Armstrong (2012), algorithms partition the data into more than one cluster and perform multiple regression of the data within each cluster in order to estimate relationships between variables.
Conclusion
Introduction
Address: Was used to understand the company's location in relation to the residential address. Property profile: includes when the property has sold overtime and the property's price. Prior to the data modelling, extreme values of the property price (typically flagging farms, larger commercial properties) have been excluded from the data set entered into the model.
Research methodology and design
Research method
The main model developed was based on three independent variables (household profile, asset profile and wealth profile). The framework is intended to explore the relationship between the three independent variables and the dependent variable, which is the decision to sell the home. This has been measured by objective means (e.g. increase in product use, growth in use), justifying a positivist approach.
Design
Create a small database based on the population data provided, covering the different data categories. A comprehensive literature review on the conversion of predictive models to business intelligence that would improve business decisions. A propensity sales model would be built for the real estate industry to demonstrate how data-driven results can influence marketing decisions.
Quantitative method
45 hypothesis analysis is expressed numerically and usually by statistical means (Creswell, 2003). During the construction of the data in the preparation of the model, an outline of the independent and dependent variables with measures and size would be captured. If not, then the propensity models would be re-run until all non-significant variables were removed.
Location of the study
The final step was to draw preferred outcomes for real estate agents by implementing.
Target Population
Outliers
To improve the quality of the data being modeled, outliers are excluded from the analysis so that they do not distort the results. Excluding outliers would improve data integrity because outliers may be related to incorrect data recording. A total of 13 properties had a purchase price of more than one billion Rand and are excluded from the dataset.
Research instruments
Secondary data
Date of Deceased: Potentially indicates that the deceased's assets will be disposed of, which includes the sale of assets. By expressing the property price per square meter, the model will inform about the relative price of the property. Perhaps the wealthier position of the higher LSM 8,9,10 allows them to sell their properties more easily.
The regression model further indicates that the price per square meter and age of the property are the most important variables. The key findings in this study will be discussed in the context of the research hypothesis, which was indicated in Chapter 1. Consumer characteristics that corroborated this relationship were evident in the consumer's affordability in terms of income groups, LSM, age of the consumer and the property lifespan (age of the property).
These factors were not considered in this study and would potentially be significant contributors to improving model accuracy.
Data collection
Data flow
Data categories
- Credit Bureau – Consumer Data
- CIPC – Directorship Data
- Deeds – Property Data
- Home Affairs – Personal Data
Data analysis
Data extract and cleanup process
Ethical Considerations
To obtain ethical approval for this study, the researcher first obtained a gatekeeper from the Managing Director, Mr. Once the gatekeeper letter was received, the research proposal was submitted to the University of KwaZulu-Natal ethics committee for study approval. Both the gatekeeper letter from Metonymy (Pty) Ltd and the approval letter from the Ethics Committee can be found in the appendix to this research report.
Conceptual Framework – Scorecard Model
Key attributes of the propensity to sell model
Life Cycle: This refers to property ownership, which is defined by the length of time in months from the time the property is purchased to the time it is sold. If the value of the variable = 1, then the property has been sold, otherwise 0 means it belongs to the current owner. A present value formula using the 5-year average CPI rate is applied to the "Purchase Amount".
Key attributes to measure include
- Household profile
- Property profile
- Wealth profile
Younger people related to the current size of the property as well as related to the age of the bond was a good indicator if you were to sell e.g. Sales rate of properties in this suburb: assumes outflow of properties that have been classified into properties with similar properties, e.g. Linking back to the company's registration date would show how old the company is, giving insight into the seniority status of an individual as a director.
Regression analysis
Conclusion
Introduction
Source of data
Variable selection
Data cleaning and preparation
Outliers/Missing Data
68 marital status, "erf size" or "house size" etc.: all such cases with missing criteria in the list of selected variables were omitted from the dataset.
Data manipulation
Transformation
When the nominal values are dummy coded into a binary variable with 1 if present, 0 otherwise, attribute categories that are not significant can be manually removed from the model. The following attributes should be dummy coded to a 1 if the attribute was present, otherwise to 0. Replacing missing values was performed in R because R can handle nominal attributes and not missing data.
Testing of Assumptions - Binary Logistic Regression
Keita (2012) stated that, as part of the definition of consumer behavior in Chapter 2, a consumer's choice and purchasing behavior can be described as utility maximization, subject to budget constraints. That is, according to the model summary, the following variables have been found to be significant predictors of sales: property price per square meter, the age of the property, age, gender, LSM and credit risk scores of the homeowners. Although marital status in the study remained neutral in terms of the statistical results, changes in family structure were expected to inform housing decision-making.
The model summary proved that the following variables were significant predictors of sale: property price per square meter, the age of the property, age, gender, LSM and credit risk scores of the homeowners. However, given the data used and the statistical test applied to determine its suitability, it provides a high probability of confidence that the model is a good approximation of the data.
Data exploration
Sale X Property Age
Sale X Property Owner Age Group
Sale X Property Owner Gender
Sale X Living Standard Measurement (LSM)
Sale X Income Band
Sale X Credit Risk Score
Sale X Director
Sale X Deceased Indicator
Sale X Marital Status
Sale X Property Type
The Logistic Regression Model
For example, the gender variable, which has 2 levels: male and female, was converted into 2 separate variables male and female, each of the form 0 or 1 (yes or no). The train subset was used to build the predictive model, while the test dataset was used to evaluate the accuracy of the model. The model with the lowest AIC score was selected and insignificant variables were removed from the model.
Model validation
Customer segmentation
Contribution of the model to business strategy within the real estate industry
Based on Verma's (2012) comment that an individual's decision-making is based on the search for information and alternative evaluations, this behavior can be adopted into a business strategy. For example, given the findings that older people generally sell their properties within a certain age bracket, this can be a target market for real estate agents in two forms, viz. a real estate business strategy must understand this behavior in order to leverage marketing efforts. .
Summary
Introduction
Key findings
Research objectives
Result: H1a is not rejected as the study proved that it is possible to predict real estate sales using logistic regression with 73.4% accuracy. Result: We fail to reject H1b as the study used title deed data and demographic information of homeowners to build a regression model that can predict property sales with 73.4% accuracy. Result: We fail to reject H1c as the study proved that it is possible to predict property sales using logistic regression with an accuracy of 73.4% based on 3 popular KwaZulu-Natal suburbs.
Implications of the findings
Recommendations emerging from the study
Recommendations for future studies
Limitations of the research
Conclusion
Online] Available at:
GATEKEEPER’S LETTER
ETHICAL CLEARANCE
TURNITIN REPORT
