PP2 Slides

You might also like

Download as pptx, pdf, or txt
Download as pptx, pdf, or txt
You are on page 1of 137

LOCATION INTELLIGENCE AND MACHINE

LEARNING BASED E-COMMERCE PLATFORM


FOR RESIDENTIAL REAL-ESTATE INDUSTRY
TEAM

Supervisor – Prof. Samantha Thelijjagoda


Co-Supervisor – Ms. Buddhima Attanayake
External Supervisor – Mr. Lanil Marasinghe (Senior Software Engineer -Sysco LABS)

STUDENT ID NAME
IT18408584 Balasooriya R.P.T.U
IT19122656 Kavishka K. A. L
IT19033006 Dananjana W.G.W.M .I
IT19055404 Shafi M.Y.B

2
INTRODUCTION

Background

Research problem

Research objectives

Overall system diagram


BACKGROUND
•Nature of real estate industry​
 High initial capital required
 Low liquidity

•Importance of decision making in residential industry


 Affects future lifestyle of the buyers​
 Lot of people spent high cost and time effort to make residential choices

•Why we need inCarto ?


 Powered by location intelligence and machine learning
 Helps to make better residential choices

4
RESEARCH PROBLEMS

 How to minimize the financial risk in investing / purchasing


real estate​

 How to minimize the high cost and time effort required to make a
proper real estate decision​

 How to guide buyers with lack knowledge about the


location characteristics​

5
RESEARCH PROBLEMS
 How to guide the buyers with lack knowledge about the domain
to choose the most optimal alternative​

 How to guide the sellers with lack knowledge about property


marketing and attract buyers​

 How to predict price fluctuation of the properties ​

6
MAIN OBJECTIVE

 Develop an E-commerce platform for residential real


estate industry which can help users to make wise and financially
safe decision , with the use of location intelligence and
machine learning

7
SUB OBJECTIVES

01 02 03 04

Provide location analytics Personal preferences- Machine learning and PRICE PREDICTION
and provide quality index based suitability score image processing OF LANDS BASED
framework for calculation and based extensive seller ON HISTORICAL
check location quality​ alternative suggestion support system DATA

8
SYSTEM DIAGRAM

9
COMMERCIALIZATION
ASPECT
Why should buyers use inCarto ?

Why should sellers use inCarto ?

inCarto APIs
WHY SHOULD BUYERS USE INCARTO

• Can get deep insights regarding the location characteristics


• Helps to make financially safe decisions
• Helps to make decisions based personal preferences
• Trustworthy properties , minor chance caught into frauds

11
WHY SHOULD SELLERS USE INCARTO

• Seller journey will be smooth due to extensive seller support


system
• Can expect larger pool of buyers
• Properties will be automatically suggested appropriate buyers

12
INCARTO API

• Third party real estate sellers can integrate inCarto API with their
platforms

13
Government organizations and big real estate companies can
use insights provided by inCarto for proper residential project
planning .

14
Overall location Quality index analyzer and analytics provider
IT19033006|Dananjana
Software Engineering
W.G.W.M.I

IT19033006| Dananjana W.G.W.M.I | 2022-013


Background

INTRODUCTION Research Gap

Research Problem

Specific Objectives

16
IT19033006| Dananjana W.G.W.M.I | 2022-013
BACKGROUND

• What is location Intelligence ?

• Why location intelligence for residential real estate industry?

• How inCarto assists in making efficient and broad spatial


decisions using location intelligence

1
IT19033006| Dananjana W.G.W.M.I | 2022-013 7
RESEARCH GAP
Provide Provide basic Provide Provide Provides Provides Provide
basic info info with the location location insights on insights on overall
location nearby analytics transport neighborhoo location
properties facilities d feasibility for
residence

ikman.lk ✔ ╳ ╳ ╳ ╳ ╳ ╳

LankaProperty ✔ ✔ ✔ ╳ ╳ ╳ ╳
Web
Prime Lands ✔ ✔ ╳ ╳ ╳ ╳ ╳

Lankaland.lk ✔ ╳ ╳ ╳ ╳ ╳ ╳

MyHome.ie[2] ✔ ✔ ✔ ╳ ╳ For some exte


nd

591.com.tw[1] ✔ ✔ For some exte


nd
╳ For some
extend
╳ ╳

1
IT19033006| Dananjana W.G.W.M.I | 2022-013 8
RESEARCH GAP
• Environmental information, neighborhood status and safety are not widely represented
in the existing websites[3].

• Real-estate websites only focus on the positive criteria of a location[5].

• Most maps and indices lack a metric and only present the locations of different
facilities and values[5].

• Existing systems not represent the analytics based on location factors.


1
IT19033006| Dananjana W.G.W.M.I | 2022-013 9
RESEARCH PROBLEM

• How to transform location data to into a usable output to


support efficient and wide spatial decision making.

• How to provide accurate location insights.

2
IT19033006| Dananjana W.G.W.M.I | 2022-013 0
SPECIFIC OBJECTIVES
•Provide location indexes based on
location factors
MAIN OBJECTIVE
•Provide indexes based on census data
• Developing a tool to •Allocate weights using common
provide location quality parameters and behaviors
indexes and related •Provide analytics based on collected
insights/analytics using data.
location intelligence

SUB OBJECTIVES

IT19033006| Dananjana W.G.W.M.I | 2022-013 2


1
Requirements

System Diagram

Technologies
METHODOLOGY
Completion of project

WBS

Gantt Chart

2
IT19033006| Dananjana W.G.W.M.I | 2022-013 2
REQUIREMENTS
Functional Requirements Non-Functional Requirements
• User should be able to choose or select a • Availability
location • Reliability
• User should be able to get indexes based on • Performance
location factors • Security
• Allocate accurate weights for location • Usability
factors
• User should be able to get location insights
23

23
IT19033006| Dananjana W.G.W.M.I | 2022-013
24
TECHNOLOGIES

• Node JS
• React and Redux
• Mongo DB
• Google MAPS API
• AWS

25 25
IT19033006| Dananjana W.G.W.M.I | 2022-013
PROJECT COMPLETION

26 26
IT19033006| Dananjana W.G.W.M.I | 2022-013
Calculating Values for Quality Indexes
Location Quality indexes Location factors considered
Health care quality index Hospitals(general/private) , pharmacies , dispensaries , dental centers ,
Veterinary Care etc ..
Education quality index Primary schools , Secondary schools , Universities/Higher education centers

Essential Good accessibility index Supermarkets , banks/ATMs , Hotels/restaurants , Clothing / shoe


centers/salon , gas stations ect .

Recreational facilities index Gyms/ fitness centers , park, walking areas , religious places , movie theaters
Transport index Bus stations , train stations , taxi stations , parking plots , airports etc .
Environmental Quality Index Rivers , mountains , hills , Forests etc

• For each Index, a value between 0 and 10 is calculated using defined location quality algorithms.

27
HEALTH CARE QUALITY INDEX
Within 5km radius from a selected location
Common weights based on Count (if) 0 1-2 2-4 4-5 5<
location factor counts
Weight 0/10 (wc1) 3/10(wc2) 5/10(wc3) 8/10(wc4) 10/10(wc5)

Independent Weights based on common user patterns and behaviors


location factor
Hospitals Allocated weight = 8 (most needed healthcare facility for
consumers) (wi1)
Pharmacies Allocated weight = 5 (people often visit pharmacies ) (wi2)
Dispensaries Allocated weight = 3 (people visits dispensaries when
comparing to dental centers but less than pharmacies ) (wi3)
Dental Centers Allocated weight = 2 (people do not often visit dental centers)
(wi4)

Other (ex. Allocated weight = 1(less needed) (wi5)


Veterinary Care)
HEALTH CARE QUALITY INDEX
Range Description

Location factor count 0–4 The selected area has less healthcare facilities

Hospitals  wi1 5  wc4


Pharmacies wi2 3 wc3 4- 7 The selected area has average healthcare
facilities
Dental centers wi3 2wc2
7-10 The selected area has good healthcare facilities
Dispensaries  wi4 2wc2
Other  wi5 0 wc1

Max Total (Tmax) = 200

Max Weight (Wmax) = 10

Health care quality Algorithm =


(wi1wc4 + wi2wc3 +wi3wc2+wi4wc2+wi5wc1) /(Tmax/ Wmax)

29
EDUCATION QUALITY INDEX
Within 5km radius from a selected location
Common weights based on
location factor counts Count (if) 0 1-2 2-4 4-5 5<
Weight 0/10 (wc1) 3/10(wc2) 5/10(wc3) 8/10(wc4) 10/10(wc5)

Independent Weights based on common user


location factor patterns and behaviors
Primary Allocated weight = 3 (secondly most
schools concerned education factor when
selecting a new property ) (wi1)

Secondary Allocated weight = 5 (most concerned


schools education factor when selecting a new
property) (wi2)

University/ Allocated weight = 2 (people have less


Higher concern on nearby universities when
education comparing to schools) (wi3)
centers
EDUCATION QUALITY INDEX Range
0–4
Description
The selected area has less education
facilities

4- 7 The selected area has average


Location factor count education facilities

Primary schools  wi1 5  wc4


7-10 The selected area has good education
facilities
Secondary schools wi2 3 wc3
universities wi3 2wc2

Max Total (Tmax) = 100

Max Weight (Wmax) = 10

Education quality algorithm =


(wi1wc4 + wi2wc3 +wi3wc2) /(Tmax/ Wmax)

31
ESSENTIAL GOOD ACCESSIBILITY INDEX
Within 5km radius from a selected location
Common weights based on Count (if) 0 1-2 2-4 4-5 5<
location factor counts
Weight 0/10 (wc1) 3/10(wc2) 5/10(wc3) 8/10(wc4) 10/10(wc5)

Independent Weights based on common user patterns


location factor and behaviors
Supermarkets Allocated weight = 8 (highly concerned) (wi1)
Banks /ATMs Allocated weight = 5 (highly concerned but
less than supermarkets) (wi2)
Hotel/ Allocated weight = 3 (moderately
Restaurants concerned ) (wi3)
Clothing / shoe Allocated weight = 2 (moderately concerned
centers/salon less than hotels and restaurants) (wi3)

Other (ex. Gas Allocated weight = 1(less needed comparing


station) o other facilities) (wi3)
ESSENTIAL GOOD ACCESSIBILITY INDEX
Range Description
Location factor count
0–4 The selected area has less healthcare facilities
Hospitals  wi1 5  wc4
Pharmacies wi2 3 wc3 4- 7 The selected area has average healthcare
facilities
Dental centers wi3 2wc2
Dispensaries  wi4 2wc2 7-10 The selected area has good healthcare facilities

Other  wi5 0 wc1


Max Total (Tmax) = 200

Max Weight (Wmax) = 10

Essential good quality Algorithm =


(wi1wc4 + wi2wc3 +wi3wc2+wi4wc2+wi5wc1) /(Tmax/ Wmax)

33
RECREATIONAL FACILITIES INDEX
Within 5km radius from a selected location
Common weights based on Count (if) 0 1-2 2-4 4-5 5<
location factor counts
Weight 0/10 (wc1) 3/10(wc2) 5/10(wc3) 8/10(wc4) 10/10(wc5)

Independent Weights based on common user patterns


location factor and behaviors
Gym/fitness Allocated weight = 8 (highly concerned in
centers urban areas) (wi1)
Park / walking Allocated weight = 5 (highly concerned but
areas etc. less than gyms in urban areas) (wi2)
Religious places Allocated weight = 4 (moderately concerned
when comparing to other recreational
facilities) (wi3)

Other (ex. Allocated weight = 3(less needed comparing


Movie theaters ) to other facilities) (wi4)
RECREATIONAL FACILITIES INDEX
Range Description
0–4 The selected area has less recreational
Location factor count facilities
Gym/fitness centers wi1 5  wc4
4- 7 The selected area has average
Parks  wi2 3 wc3 recreational facilities

Religious places  wi3 2wc2 The selected area has good


7-10
recreational facilities
Other  wi4 0 wc1

Max Total (Tmax) = 200

Max Weight (Wmax) = 10

Recreational facility quality algorithm =


(wi1wc4 + wi2wc3 +wi3wc2+wi4wc2) /(Tmax/ Wmax)

35
TRANSPORT INDEX
Independent Common weights based on location factor
location factor count(Within 5km radius from a selected
location
)
Bus stations (wi1) If count >= 3  weight = 70 (wcx)
Else  weight = 10 (wcy)
Train stations(wi2) Additional 10 (wcy) will be added if count >= 2
Else 0
Parking plots(wi3) Additional 10 (wcy) will be added if count > 3
Else 0

Other places(ex. Additional 10 (wcy) will be added if count > 1


Airports ) (wi5) Else 0
TRANSPORT INDEX Range Description
0–4 The selected area has less transport
facilities
Location factor count
Bus stands  wi1 5  wcx 4- 7 The selected area has average transport
facilities
Train stations wi2 3 0
7-10 The selected area has good transport
Parking  wi3 2wcx facilities
Other  wi4 0 0

Max Total (Tmax) = 200

Max Weight (Wmax) = 10

Transport quality algorithm =


(wi1wcx + 0 +wi3wcx+0) /(Tmax/ Wmax)
ENVIRONMENTAL QUALITY FACTOR
Independent location factor Availably(Within 0.5km
radius from a selected
location
)
Rivers (Floods are more of If count >= 1  yes
common occurrence in Sri Lanka Else  no
than the other natural disasters.)

Mountains (Heavy rains and If count >= 1  yes


geological changes in the hill Else  no
country, mountainous areas, have
increased the occurrences of
frequent landslides)

Forests (animal attacks ect .) If count >= 1  yes


Else  no
FRONT-END IMPLEMENTATION

Quality Index Framework

39
FRONT-END IMPLEMENTATION

Map- 5km radius view and nearby properties

40
FRONT-END IMPLEMENTATION

Map- map view with live traffic option


41
FRONT-END IMPLEMENTATION

Map with location street view

42
FRONT-END IMPLEMENTATION

Location analytics

43
PROJECT COMPLETION
Extract location data from google API 100%
Define quality algorithms 100%
Allocate weights based on selected property counts radius 100%
Accurate data using common patterns and behaviors 100%

Generate indexes based on location 100%

Provide analytics based on quality algorithm outputs 100%

Advanced dashboard(provide analytics) 100%

44
WBS
Overall location Quality index analyzer and
analytics provider​

1. Quality index 2.Quality index algorithm


algorithm based on 3.Show location insights
based on census data
location factors

1.1 extract data from 2.1 extract census data 3.1 Get outputs from
google API from available databases quality index algorithms
1.2 Accurate using
3.1 show analytics based
common patterns and
2.2 Allocate weights on location
behaviors

1.3 Allocate weights 3.3 Advanced dashboard


2.3 Generate indexes

1.4 Generate indexes

45
IT19033006| Dananjana W.G.W.M.I | 2022-013
GANTT CHART

46
IT19033006| Dananjana W.G.W.M.I | 2022-013
REFERENCES
[1]A. T. Co., "newhouse.591.com," Addcn Technology Co., 2007. [Online]. Available:
https://newhouse.591.com.tw/housing-list.html?rid=1&sid=12.

[2]MyHome.ie, "www.myhome.ie," 2021. [Online]. Available: https://www.myhome.ie/. [Accessed 2021].


MyHome.ie - Houses, Apartments, Irish Property for Sale in Ireland

[3]“National Output, Expenditure and Income,” 2021.


[4]“Real Estate Definition.” https://www.investopedia.com/terms/r/realestate.asp (accessed Jan. 24, 2022).
[5]H. Rabiei-Dastjerdi, G. McArdle, S. A. Matthews, and P. Keenan, “Gap analysis in decision support systems
for real-estate in the era of the digital earth,” International Journal of Digital Earth, vol. 14, no. 1, pp. 121–138,
2021, doi: 10.1080/17538947.2020.1808719.

47
IT19033006| Dananjana W.G.W.M.I | 2022-013
Personal preferences-based suitability analysis and
alternative recommendation

IT18408584 | Balasooriya R.P.T.U

Software Engineering

IT18408584 | Balasooriya R.P.T.U | 2022-013


Background

INTRODUCTION Research Gap

Research Problem

Specific Objectives

49
IT18408584| Balasooriya R.P.T.U | 2022-013
BACKGROUND
• What is suitability Analysis
• Process of qualify, compare, and rank candidate sites based on how closely they adhere to criteria that you
select and define.

• What is suitability score ?


• Why personal preferences-based suitability analysis is important to
residential real estate industry
• Why personal preferences-based alternative recommendation is important
• Alternative recommendations is desired among decision markers in the real-estate domain .

50
IT18408584| Balasooriya R.P.T.U | 2022-013
RESEARCH GAP
System Random location Previous search- Personal Personal preferences
recommendation based location preferences-based based alternative
recommendation suitability analysis suggestion

Ikman.lk No No No No
Homelands No No No No
Prime lands Yes No No No
WP Rupasinghe No Yes No No
Realtee.com No Yes No No
FSBO.com No Yes No No
Zillow.com No Yes No No

51
IT18408584| Balasooriya R.P.T.U | 2022-013
RESEARCH GAP

• GIS Tools :Arc GIS and Q GIS

52
IT18408584| Balasooriya R.P.T.U | 2022-013
RESEARCH GAP

• Higher domain knowledge required

• User should define the criteria

• User should have the data set

• Criteria is not dynamic

• Cannot be understand be by any nontechnical person


53
IT18408584| Balasooriya R.P.T.U | 2022-013
RESEARCH QUESTIONS

• How to implement personal preferences-based location


suitability scoring mechanism ?

• How to suggest optimal alternative recommendation considering


both price and suitability aspects ?

54
IT18408584| Balasooriya R.P.T.U | 2022-013
SPECIFIC OBJECTIVES • Research the criteria which should be
included to preference profile
MAIN OBJECTIVE • Design and implement personal
preference-based suitability scoring
• Design web-based solution to where algorithm
users check how far a selected • Design and implement alternative
residential location is matches with suggestion algorithm
personal preferences , and find about
are there any better alternatives
SUB OBJECTIVES

5
5
System Diagram

Technologies
METHODOLOGY
Requirements

WBS

Gantt Chart
56
IT18408584| Balasooriya R.P.T.U | 2022-013
57
TECHNOLOGIES

• Node JS
• React and Redux
• Mongo DB
• Google MAPS API
• AWS

58
5
IT18408584| Balasooriya R.P.T.U | 2022-013 8
REQUIREMENTS
Functional requirements Nonfunctional requirements
• User should be able save the personal • Non steep learning curve
preferences as a profile • Accuracy
• User should be able to change • Performance
preferences • Security
• User should be able to see the • Usability
suitability level as score
• User should be able compare with
alternatives
59
IT18408584| Balasooriya R.P.T.U | 2022-013
COMPLETION OF THE PROJECT

60
CRITERIA SELECTION FOR SUITABILITY ANALYZING
ALGORITHM
Main Criteria Sub criteria
Accessibility to Essential Goods Supermarkets , Grocery stores, Hardware Stores
Educational Facilities Pres schools , Non state schools , State Government
schools , State Universities , Private Universities ,

Health Care Facilities Hospitals , Pharmacies , Dental clinics or dentists


Recreational Sporting Facilities Gym and Fitness centers , Jogging tracks ,
Playgrounds , Swimming pools , Movie theaters &
Cinema

Transport Facilities Bus stops , Railway stations , Highways and


Expressways

Religious and cultural centers Buddhist temple , Church , Mosque , Hindu temple

61
EVALUATING SUB CRITERIA

Minimum expected travel time Maxt


Maximum expected travel time Mint
Actual travel time Actt
Sub criteria full marks Sf
Subcriteria score Ss

62
EVALUATING SUB CRITERIA
Actt < Mint :
Actt > Maxt :
Sub criteria score = Sf
Sub criteria Score = 0

Mint < Act < Maxt :

Sub criteria score = sf –sf * (Act – Mint ) / (Maxt – Mint)

63
EVALUATE MAIN CRITERIA

Main Criteria Score(MS) = (∑ Ss ) / (Σ Sf )

64
FINAL SCORE

Final Score (MS) = (Σ Msi *Wi ) / (Σ Wi)

65
ALTERNATIVE REOMMENDATION

Budget feasibility Score

Budget Expectation

Alternative Locations
Alternative Preference feasibility Score
Preference Expectation
recommendation system

Total Weighted feasibility


Location characteristics

66
USER INTERFACES

67
USER INTERFACES

68
COMPLETION OF PROJECT
Finalizing main and sub criteria for scoring algorithm 100%
Design and Implement scoring algorithm 100%
Backend service implementation for profile management 100%
and suitability scoring
Preference profile frontend and score view frontend 100%
implementation
Implement alternative suggestion algorithm 100 %
Implement advance statistics visualizations to select 100 %
optimal alternatives

69
BACKEND REST API

70
WORK BREAK DOWN STRRUCTURE
Personal preference-based suitability analysis and
alternative recommendation

1 .Analyze personal 2 .Calculate Suitability 3.Alternative


preference profile Score recommendation

1.1Identify selected 2.1 Check travel time 3.1 Analyze suitability


main criteria using Google MAPS aspect

1.2Identify selected 3.2 Analyze budget


sub criteria 2.2 Allocate marks
aspect

1.3Identify 3.2 Suggest optimal


2.3 Calculate sum
prioritization alternatives
71
IT18408584| Balasooriya R.P.T.U | 2022-013
GANNT CHART

72
IT18408584| Balasooriya R.P.T.U | 2022-013
REFERENCES
[1]“Real Estate - Industry Overview, Types of Real Estate, Careers,” corporatefinanceinstitute.
https://corporatefinanceinstitute.com/resources/careers/jobs/real-estate/ (accessed Jan. 23, 2022).
[2]“How To Start A Real Estate Business In 2022 | FortuneBuilders.” https://www.fortunebuilders.com/a-beginners-
guide-to-starting-a-real-estate-business/ (accessed Jan. 24, 2022).
[3]“National Output, Expenditure and Income,” 2021.
[4]“Real Estate Definition.” https://www.investopedia.com/terms/r/realestate.asp (accessed Jan. 24, 2022).
[5]H. Rabiei-Dastjerdi, G. McArdle, S. A. Matthews, and P. Keenan, “Gap analysis in decision support systems for
real-estate in the era of the digital earth,” International Journal of Digital Earth, vol. 14, no. 1, pp. 121–138, 2021, doi:
10.1080/17538947.2020.1808719.

73
IT18408584| Balasooriya R.P.T.U | 2022-013
Predict price fluctuation for land locations by using machine learning.

IT19122656|Kavishka K.A.L

Software Engineering

IT19122656 | Kavishka K.A.L | 2022-013


Background

INTRODUCTION Research Gap

Research Problem

Specific Objectives

IT19122656 | Kavishka K.A.L | 2022-013 75


IT19122656 | Kavishka K.A.L | 2022-013
BACKGROUND
• Real estate property price prediction gained an important role in
current Sri Lankan society mainly to the buyer and sellers in the
field of marketing.

• The real estate property prices are changed rapidly during


recent past years.

IT19122656 | Kavishka K.A.L | 2022-013 76


RESEARCH GAP
System Current Price fluctuation Historical price change view
prediction system chart
Ikman.lk

Homelands

Lanka Property web

Prime Lands

Lankaland.lk

W.P. Rupasinghe & Co

IT19122656 | Kavishka K.A.L | 2022-013 77


RESEARCH QUESTIONS

• Why price prediction is important to the buyer


and seller?

IT19122656 | Kavishka K.A.L | 2022-013 78


• Create datasets by using of
SPECIFIC OBJECTIVES historical datasets.
• Testing and training datasets.
MAIN OBJECTIVE • Pre-process data and features
engineering .
Current price prediction • Create the model.
Historical price change indicator
• Create a function to get user
selected locations as an input.

SUB OBJECTIVES

IT19122656 | Kavishka K.A.L | 2022-013 7


9
System Diagram

Technologies
METHODOLOGY
Requirements

WBS

Gantt Chart
IT19122656 | Kavishka K.A.L | 2022-013 80
SYSTEM DIAGRAM

81
TECHNOLOGIES
Techniques :
• sklearn framework
• Linear regression

Tools:
• Python
• VS code
• Jupyter notebook
• FLASK
• NumPy
82
IT19122656 | Kavishka K.A.L | 2022-013
WORK BREAKOUT STRUCTURE

IT19122656 | Kavishka K.A.L | 2022-013 83


GANTT CHART

IT19122656 | Kavishka K.A.L | 2022-013 84


COMPLETION OF THE PROJECT

85
DATASET
2021 NEWEST DATASET

86
HISTORICAL DATASET

87
Uncleaned Dataset

Cleaned Dataset

88
TRAINING THE PRICE PREDICTNG MODEL
Algorithm

89
90
91
USER INTERFACES

92
HISTORICAL PRICE FLUCTUATION CHART

93
PROJECT COMPLETION
Current price prediction – 100%

Current price prediction 100%


Gather newest dataset (2021 dataset) 100%
Clean dataset 100%

Train the model 100%

Implement and integrate frontend and backend 100%


Gather Historical dataset 100%
Yearly price fluctuation chart using historical data set 100%

94
REFERENCES
[1]CBSL, “Land Valuation Indicator1 – First Half of 2021 ” , [Online] .[Available:
https://www.cbsl.gov.lk /sites /default / files/cbslweb_documents/press/pr/press_20210806_land_valuation_indicat
or_first_half_2021_e.pdf
, Accessed 31-Jan-2022

[2] J. R. Quinlan, “Combining instance-based and model-based learning,” Morgan Kaufmann, 1993, pp. 236–243.

[3] PCVC, “Review of Price Formation of Land and Landed Properties in Colombo: Is it a Myth or Reality?”, [Online].
[Available:
https://pcvc.lk/wp-content/uploads/2016/12/Land-Pricing-formation-in-Colombo-published-on-15-12-2016.pdf ,
Accessed 31-Jan-2022

IT19122656 | Kavishka K.A.L | 2022-013 95


Machine Learning and Image Processing Based Extensive Seller Support System

IT19055404|Shafi M. Y. B.
Software Engineering

IT9055404 | Shafi M. Y. B | 2022-13


Background

INTRODUCTION Research Gap

Research Problem

Specific Objectives

9
IT9055404 | Shafi M. Y. B | 2022-13 7
BACKGROUND

Listing a property involves

• Posting photos of the property


• Creating a good property description

What is the importance of the property description and images in real


estate property advertising?

9
IT9055404 | Shafi M. Y. B | 2022-13 8
BACKGROUND

Why would a seller support system with automated property


description generation, automated image tagging, and image
enhancement be useful?

9
IT9055404 | Shafi M. Y. B | 2022-13 9
BACKGROUND

When listing properties,

1. Creating a property description and annotating property photos can


be time consuming and requires manual effort.
2. Some written property descriptions do not include important
special features of a property.
3. False location details can be entered
4. Posting low quality images
1
IT9055404 | Shafi M. Y. B | 2022-13 0
0
BACKGROUND

Description without detailed


information

A quality description with important details included


1
IT9055404 | Shafi M. Y. B | 2022-13 0
1
BACKGROUND

Good quality images

Low quality images


1
IT9055404 | Shafi M. Y. B | 2022-13 0
2
RESEARCH GAP
Scene Tagging Image Scene tagging Property Description gene
(Manual) enhancement (automated) description ration
generation with with location
property images details
ikman.lk

LankaPropertyWeb

Prime Lands

Lankaland.lk

1
IT9055404 | Shafi M. Y. B | 2022-13 0
3
RESEARCH PROBLEM

How to automate the real estate property image annotating and


property description generation, to increase the efficiency of
the property listing process and to attract more buyers, with the help
of image processing and computer vision?

1
IT9055404 | Shafi M. Y. B | 2022-13 0
4
SPECIFIC OBJECTIVES

MAIN OBJECTIVE
• Image enhancement to
enhance the input images
Design computer vision based,
seller support system for • Image classification for scene
automating real estate image tagging
tagging and property description • Obtain location details from
generation google map API
• Create property description

SUB OBJECTIVES

IT9055404 | Shafi M. Y. B | 2022-13 1


0
System Diagram

Technologies
METHODOLOGY
Requirements

WBS

Gantt Chart
1
IT9055404 | Shafi M. Y. B | 2022-13 0
6
SYSTEM DIAGRAM


​ IT9055404 | Shafi M. Y. B | 2022-13
107
TECHNOLOGIES
• Python
• Google Colab
• Google Maps API
• Tensorflow
Algorithms and Techniques
• CLAHE (contrast-limited adaptive histogram equalization)
• EfficientNetB5 model

1
IT9055404 | Shafi M. Y. B | 2022-13 0
8
REQUIREMENTS
Functional requirements Non-Functional requirements
•User should be able to enhance the
image if needed • Accuracy
•Image label should be auto generated • Usability
when user uploads the image
•User should be able to generate the • Reliability
description when photos of property is • Availability
input
•User should be able to edit the
property description details if needed
•Location details description should not
be allowed to change by seller

​ IT9055404 | Shafi M. Y. B | 2022-13
1
0
9
IMAGE ENHANCEMENT MODULE

CLAHE ENHANCEMENT

110
HISTOGRAM EQUALIZATION VS CLAHE(CONTRAST
LIMITED ADAPTIVE HISTOGRAM EQUALIZATION)

111
IMAGE CLASSIFICATION MODULE
DATA SET FOR IMAGE CLASSIFICATION

80% data for Training


20% data for Testing

112
SAMPLE IMAGES FROM THE DATA SET

113
114
IMAGE PREPROCESSING
IMAGE AUGMENTATION

115
FINE TUNING TRANSFER LEARNING MODEL

116
TRAINING MODEL

117
MODEL ACCURACY AND LOSS GRAPH

118
PREDICTION RESULTS

119
120
CONFUSION MATRIX

121
PROPERTY DESCRIPTION GENERATION MODULE
DATASET

• Flickr 8k dataset
• 6000 images for training
• 1000 images for testing
• Training dataset vocabulary size - 7507

122
IMAGE CAPTION GENERATION MODEL
MODEL SUMMERY

123
IMAGE CAPTION GENERATION MODEL
MODEL TRAINING

124
IMAGE CAPTION GENERATION MODEL TESTING RESULTS

125
USER INTERFACE

126
GANTT CHART

IT19122656 | Kavishka K.A.L | 2022-013 127


WORK BREAKOUT STRUCTURE

IT19122656 | Kavishka K.A.L | 2022-013 128


PROJECT COMPLETION
Image Enhancement 100%
Image classification Model creation 100%
Create Dataset 100%

Train the model 100%

Frontend Implementation 90%


Property description generation 85%

129
REFERENCES
[1] M. J. Seiler, P. Madhavan, and M. Liechty, "Toward an understanding of real estate
homebuyer internet search behavior: an application of ocular tracking technology". Journal of
Real Estate Research, vol. 34, no. 2, pp.211-242, Jan 2012.
[2] J. H. Bappy, J. R. Barr, N. Srinivasan, and A. K. R. Roy-Chowdhury "Real estate image
classification." 2017 IEEE Winter Conference on Applications of Computer Vision (WACV).
IEEE, March 2017.
[3] D. Koch, M. Despotovic, S. Leiber, M. Sakeena, M. Döller and M. Zeppelzauer, 2020.
Real Estate Image Analysis-A Literature Review (preprint).
[4] L. Shen, and S. Ross. "Information value of property description: A machine learning
approach." Journal of Urban Economics 121 (2021): 103299.
[5] https://keras.io/api/applications/

1
IT9055404 | Shafi M. Y. B | 2022-13 3
0
BEST PRACTICES
PROPER SOLUTION STRUCTRE

132
• Proper variable naming

• Commenting

• Write code in reusable manner

133
RISK MITIGATION
• Short and manageable iterations
• Effective communication and clear agreements
• Used project management tools

135
• Microservices architecture

136
THANK YOU!

You might also like