Missing traffic data imputation considering approximate intervals: A hybrid structure integrating adaptive network-based inference and fuzzy rough set

Abstract

The incomplete traffic data will seriously influence the application of Intelligent Transportation System (ITS). In this study, a hybrid model, combining Adaptive Network-based Fuzzy Inference System (ANFIS) and Fuzzy Rough Set (FRS), is constructed to impute missing traffic volume data. Firstly, the upper and lower parameters of fuzzification in hybrid model are optimized through the combination of FRS and least square method. Then, a five-layer structure is designed to achieve imputation process under the integration of ANFIS and FRS. In addition, the imputation process is improved and the final imputation results can be estimated based on the upper and lower approximations of missing values. Finally, experiments are conducted to validate the effectiveness of the hybrid model using three evaluation indicators: Root Mean Square Error (RMSE), Correlation Coefficient (R) and Relative Accuracy (RA), under different missing rates. Though comparing with several candidate models under random missing and continuous missing types, the imputation results show that the hybrid method produces higher imputation accuracy and better stability imputation performance under different data missing rates, especially for high missing rate. The hybrid model, combining the advantages for dealing with uncertainty in FRS and strong learning ability to outliers in ANFIS, is an effective and feasible strategy to improve imputation performance for missing traffic flow data in transportation system.