Data Imputation for Traffic State Estimation and Pre-diction Using Wi-Fi Sensors

Real-time monitoring of traffic conditions is essential to support control strategies and provide useful information to travelers. With the accelerated development in transportation management systems (TMSs), traffic data collection methods have progressed rapidly. Despite the development in the data collection systems, there is missing data due to occasional sensor damage, trans-mission error, or a low penetration rate of the probe vehicle, thereby affecting the reliability and effectiveness of the Intelligent Transportation Systems (ITS). There is a need for effective data imputation methods to ensure the integrity and quality of traffic data. Two clustering-based methods for such traffic imputations are proposed in this paper—one using k-means clustering and the other using speed bins. Mean Absolute Percentage Error (MAPE) is used as a performance efficiency index for both methods. The Speed bin method was found to be more effective with a maximum MAPE of 13.8%. The maximum MAPE observed for the k-means clustering method is 22.6%.