-
Notifications
You must be signed in to change notification settings - Fork 0
/
index.html
203 lines (173 loc) · 7.43 KB
/
index.html
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
<!doctype html>
<html>
<head>
<meta charset="utf-8">
<meta http-equiv="X-UA-Compatible" content="chrome=1">
<title>Rtb by mryap</title>
<link rel="stylesheet" href="stylesheets/styles.css">
<link rel="stylesheet" href="stylesheets/github-light.css">
<meta name="viewport" content="width=device-width">
<!--[if lt IE 9]>
<script src="//html5shiv.googlecode.com/svn/trunk/html5.js"></script>
<![endif]-->
</head>
<body>
<div class="wrapper">
<header>
<h1>Rtb</h1>
<p></p>
<p class="view"><a href="https://github.com/mryap/rtb">View the Project on GitHub <small>mryap/rtb</small></a></p>
<ul>
<li><a href="https://github.com/mryap/rtb/zipball/master">Download <strong>ZIP File</strong></a></li>
<li><a href="https://github.com/mryap/rtb/tarball/master">Download <strong>TAR Ball</strong></a></li>
<li><a href="https://github.com/mryap/rtb">View On <strong>GitHub</strong></a></li>
</ul>
</header>
<section>
<h1>
<a id="prediction-model-for-dublin-rental-price-movements" class="anchor" href="#prediction-model-for-dublin-rental-price-movements" aria-hidden="true"><span aria-hidden="true" class="octicon octicon-link"></span></a>Prediction Model for Dublin rental price movements</h1>
<p>Using Azure Machine Learning studio and R’s Caret package to explores the use of data mining techniques to analyse the trend and understand the underlying influencing factors of Ireland property rental market. </p>
<p>Transaction data from publicly available Private Residential Tenancies Board (PRTB) rent index from 2005 to 2015 were used and built prediction models using regression algorithms. </p>
<h1>
<a id="data" class="anchor" href="#data" aria-hidden="true"><span aria-hidden="true" class="octicon octicon-link"></span></a>Data</h1>
<p>Private Residential Tenancies Board (PRTB), Central Statistics Office (CSO), OECD data are extracted and concatenated together as a multi-variate dataset, with rental data as the predictable column.
<a href="http://dx.doi.org/10.7910/DVN/HRKFN2">http://dx.doi.org/10.7910/DVN/HRKFN2</a> </p>
<h1>
<a id="machine-learning-algorithms-deploy" class="anchor" href="#machine-learning-algorithms-deploy" aria-hidden="true"><span aria-hidden="true" class="octicon octicon-link"></span></a>Machine Learning Algorithms Deploy</h1>
<ul>
<li>Feature engineering</li>
<li>Exploratory Analysis </li>
<li>Boosted Decision Tree Regression </li>
<li>Ordinary Least Squares Linear regression </li>
</ul>
<h1>
<a id="feature-selection" class="anchor" href="#feature-selection" aria-hidden="true"><span aria-hidden="true" class="octicon octicon-link"></span></a>Feature Selection</h1>
<p>To construct effective features in the training data, 5 training datasets are constructed based on the same raw input data, but different additional features to each training set were added.</p>
<ul>
<li>Set A = Property Type + Location + Year + Rent features for the predicted rent</li>
<li>Set B = Property Type + Location + Year + CPI + Rent features for the predicted rent</li>
<li>Set C = Property Type + Location + Year + CPI + Price-Rent-Ratio + Rent features for the predicted rent </li>
<li>Set D = Property Type + Location + CPI + Price-Rent-Ratio + Vacancy Rates + Year + Rent features for the predicted rent </li>
<li>Set E = Property Type + Location + Year + CPI + Vacancy Rates + No. of Room + Rent features for the predicted rent</li>
</ul>
<h1>
<a id="measuring-model-performance" class="anchor" href="#measuring-model-performance" aria-hidden="true"><span aria-hidden="true" class="octicon octicon-link"></span></a>Measuring Model Performance</h1>
<p align="center">
<img src="https://github.com/mryap/rtb/blob/master/media/CnE4iSpWAAAcG7v.jpg?raw=true">
</p>
<p>On a split training (70%) and test (30%) dataset, Boosted decision tree regression model is also the matching algorithm. </p>
<table>
<thead>
<tr>
<th>Split Data</th>
<th>Algorithm</th>
<th>RMSE</th>
<th>r2</th>
</tr>
</thead>
<tbody>
<tr>
<td>50/50</td>
<td>Boosted Decision Tree Regression</td>
<td>343.3298</td>
<td>0.487419</td>
</tr>
<tr>
<td>65/35</td>
<td>Boosted Decision Tree Regression</td>
<td>347.0633</td>
<td>0.473526</td>
</tr>
<tr>
<td>60/40</td>
<td>Boosted Decision Tree Regression</td>
<td>358.6932</td>
<td>0.452571</td>
</tr>
<tr>
<td>80/20</td>
<td>Boosted Decision Tree Regression</td>
<td>367.1906</td>
<td>0.411125</td>
</tr>
<tr>
<td>70/30</td>
<td>Boosted Decision Tree Regression</td>
<td>344.0009</td>
<td>0.497633</td>
</tr>
</tbody>
</table>
<p>Adding additional features like Number of Bedrooms to the data on the same training split – 70/30 produce the following outcome that Linear Regression is the best model in terms of performance</p>
<h1>
<a id="evaluating-the-importance-of-the-variables" class="anchor" href="#evaluating-the-importance-of-the-variables" aria-hidden="true"><span aria-hidden="true" class="octicon octicon-link"></span></a>Evaluating the importance of the variables</h1>
<p>Kuhn’s R caret package is use to calculate variable importance. The varImp function together with lm object class produces the following outcome. </p>
<p align="center">
<img src="https://github.com/mryap/rtb/blob/master/media/Rplot.png?raw=true">
</p>
<p>In order of permutation importance scores, results from Azure ML analysis also produce a similar ranking results </p>
<table>
<thead>
<tr>
<th>Feature</th>
<th>Score</th>
</tr>
</thead>
<tbody>
<tr>
<td>PropertyType</td>
<td>201.2722</td>
</tr>
<tr>
<td>Location</td>
<td>180.1229</td>
</tr>
<tr>
<td>CPI</td>
<td>65.83267</td>
</tr>
<tr>
<td>Price.Rent.Ratio</td>
<td>65.11127</td>
</tr>
<tr>
<td>HousingStock</td>
<td>1.358679</td>
</tr>
<tr>
<td>VacancyRate</td>
<td>0.828779</td>
</tr>
<tr>
<td>NumberofBedrooms</td>
<td>0.005061</td>
</tr>
<tr>
<td>Year</td>
<td>-0.02083</td>
</tr>
</tbody>
</table>
<h1>
<a id="conclusion--future-works" class="anchor" href="#conclusion--future-works" aria-hidden="true"><span aria-hidden="true" class="octicon octicon-link"></span></a>CONCLUSION & FUTURE WORKS</h1>
<p>AirBnB being cited as an influencing factor in the Irish rental market. For the next step, Dublin AirBnB data will be used to analyse the impact on the rental price movement. </p>
<p>According to Kieran and Gerard, house prices depend on how much individuals can borrow from financial institutions, with the amount borrowed can be a proxy of their disposable income and the current mortgage rate. Hence, the inclusion of income and interest rates offer the possibility of adding characteristics to the regression model for the next iteration of this project. </p>
<p>To further improve and enhance the predictions model to a more accurate level, this project is going to benefit from a domain expert knowledge on Ireland property market.</p>
<h1>
<a id="to-do" class="anchor" href="#to-do" aria-hidden="true"><span aria-hidden="true" class="octicon octicon-link"></span></a>TO DO</h1>
<ul>
<li>[ ] Calculate correlation<br>
</li>
<li>[ ] To remove attributes with an absolute correlation of 0.75 or higher.</li>
<li>[ ] Find out outliner</li>
<li>[ ] Deploy the model as a product </li>
</ul>
</section>
<footer>
<p>This project is maintained by <a href="https://github.com/mryap">mryap</a></p>
<p><small>Hosted on GitHub Pages — Theme by <a href="https://github.com/orderedlist">orderedlist</a></small></p>
</footer>
</div>
<script src="javascripts/scale.fix.js"></script>
</body>
</html>