City: |
Hong Kong, Hong Kong, China
|
Organization: |
Department of Real Estate and Construction, The University of Hong Kong, Hong Kong
|
Project Start Date: |
2013 (since research team working, as mentioned in the paper)
|
Project End Date: |
28 October 2018 (the article is accepted by the Resources, Conservation & Recycling Journal)
|
Reference: |
Weisheng Lu, 2019. Big data analytics to identify illegal construction waste dumping: A Hong Kong study. Resources, Conservation & Recycling Journal, February, Volume 141, pp. 264-272. DOI: https://doi.org/10.1016/j.resconrec.2018.10.039
|
Problem: |
- In Hong Kong, Environmental Protection Department (EPD) statistics from 2015 showed that 27.8% of total solid waste was generated from the construction industry and that is from all the waste that ends up in the landfill. However, despite the policies and schemes like the Construction Waste Disposal Charging Scheme (CWDCS), which regulates the contractor for proper waste disposal, the illegal dumping trend is increasing because of cost and time concerns or convenience.
- This illegal dumping is causing server environmental problems especially to wetland fauna and mangroves and takes a heavy amount of governmental budget to clean this dump. The purpose of this study was to use Big data analytics to quantify the magnitude of illegal dumping and workout countermeasures accordingly.
|
Technical Solution: |
To develop red-flag indicators, illegal dumping behavior is characterized by adopting a mixed-method approach.
- 1.A list of indicators for predicting illegal dumping activities is presented.
- 2.The second step is to develop the core algorithms, referred to as the Illegal Dumping Filter (IDF)
- 3.Supervised learning: Training data by evaluation and calibration
- a.Illegal dumping convictions cases were used as the experimental/target group
- b.A comparable sample used as the control group
For Illegal Dumping Filter (IDF) 10-fold cross-validation experiment was used as follows:
- Indicators and behavior are feed to form a binary rule guided by 3 conditions concluded by JRip, which is a Java version of the Repeated Incremental Pruning to Produce Error Reduction (RIPPER) was used.
- Pearson’s linear correlation coefficient model is used to filter the relevant and connected indicators.
- Weka (version 3.9) used for nonlinear models of correlations
- A decision tree concluded by J48, a Java version of the C4.5
- Function Radial Basis Function (RBF) classifier
- Meta-model Random committee
- Methods used for comparison:Pearson correlation analysis & Significance (2-tailed)
Finally, the IDF model was applied using the selected J48 method to filter the suspected illegal dumping actions from the database, with a view to understanding the overall magnitude of the illegal dumping problem.
|
Datasets Used: |
Dataset 1: Facility, Facility Code, Facility Name
- Title: Environmental Protection Department (EPD) Facility database
- Source: Construction Waste Disposal Charging Scheme (CWDCS)
- Date: January 2011 and December 2017
Dataset 2: Account, Contract No., Contract Name, Contract Sum, Address, Department, Type of Construction Work, detail of construction work, Remarks
- Title: Environmental Protection Department (EPD) Project database
- Source: Construction Waste Disposal Charging Scheme (CWDCS) A total of 27,536 construction projects
- Date: January 2011 and December 2017
Dataset 3: Facility, Transaction Ref. No., Account, Time-In, Weight-In, Vehicle No., Weight Out, Net Weight, Time Out, Chit No, Data of Transaction, Waste Depth
- Title: Environmental Protection Department (EPD) Waste Disposal database
- Source: Construction Waste Disposal Charging Scheme (CWDCS) 9,338,243 disposal records
- Date: January 2011 and December 2017
Dataset 4: Vehicle No, Permitted Gross Vehicle Weight
- Title: Environmental Protection Department (EPD) Vehicle database
- Source: Construction Waste Disposal Charging Scheme (CWDCS) containing 9863 vehicles
- Date: January 2011 and December 2017
|
Outcome: |
- This research identified 546 waste hauling trucks suspected of involvement in illegal dumping.
- Through big data analytics, previously unknown characteristics of illegal dumpers were identified: for example, they are freelance, and less patient in queuing at government waste disposal facilities.
- Although the analytical results cannot be used as evidence to prosecute suspected offenders, they offer important decision-support information for follow-up interventions to combat illegal dumping.
|
Issues that arose: |
- Even with big data analytics, there is no one-size-fits-all solution to urban crime identification fraud detection.
- Big data analytics has serious potential ethical ramifications and should be treated with caution.
- Big data analytics could lead to privacy infringement and other issues that still have no readily available theoretical explanation or practical solution.
|
Status: |
Terminated.
|
Entered by: |
October 30th,2020: Syeda Batool Zehra, sb.zehra@mail.utoronto.ca
|