A Hybrid Missing Data Imputation Method for Constructing City Mobility Indices
| Autor: | Sanaz Nikfalazar, Susan E. Bedingfield, Chung-Hsing Yeh, Hadi Akbarzadeh Khorshidi |
|---|---|
| Rok vydání: | 2019 |
| Předmět: |
050210 logistics & transportation
Fuzzy clustering Computer science Iterative method 05 social sciences Decision tree 02 engineering and technology Missing data computer.software_genre Knowledge extraction Missing data imputation 0502 economics and business Principal component analysis 0202 electrical engineering electronic engineering information engineering 020201 artificial intelligence & image processing Imputation (statistics) Data mining computer |
| Zdroj: | Communications in Computer and Information Science ISBN: 9789811366604 AusDM |
| DOI: | 10.1007/978-981-13-6661-1_11 |
| Popis: | An effective missing data imputation method is essential for data mining and knowledge discovery from a comprehensive database with missing values. This paper proposes a new hybrid imputation method to effectively deal with the missing data issue of the Mobility in Cities Database (MCD) to construct city mobility indices. The hybrid method integrates the advantages of decision trees and fuzzy clustering into an iterative algorithm for missing data imputation. Extensive experiments conducted on the MCD and three commonly used datasets demonstrate that the hybrid method outperforms other existing effective imputation methods. With the MCD’s missing values imputed by the hybrid method, and using factor analysis and principal component analysis, this paper constructs city mobility indices for 63 cities in the MCD based on the novel concept of city mobility supply and demand. The city mobility indices constructed under a hierarchical structure of mobility supply and demand indicators represent substantial city mobility knowledge discovered from mining the MCD. The proposed hybrid method represents a significant contribution to missing data imputation research. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|