| Abstract |
The rapid development of urban information techniques, mobile sensors, and artificial intelligence can help to generate solutions for transportation mode recognition (TMR). The Sussex-Huawei Locomotion-Transportation (SHL) recognition challenge organized at the UbiComp 2021 presents a large and realistic dataset with different activities. Different from the previous three years that aimed at transportation mode recognition from the motion sensors, the goal of this year is to recognize 8 modes of locomotion and transportation (activities) in a user-independent manner based on radio-data, including GPS reception, GPS location, WiFi reception, and GSM cell tower scans. In this paper, our team (We can fly) summarizes our submission to the competition. We first preprocess the data, divide the sample time window and deal with missing values, and then extract 303 features from the given sensor, and finally feed these features into the LightGBM classifier. In the experiment, we utilized the training datasets to train our model and achieved macro-F1 score of 0.665 on the valid datasets. © 2021 ACM. |
