Research on Combination Filling Method for Multi-Dimensional Traffic Accident Missing Data

At present, in the collection of traffic accident data, there are often missing data of different types and dimensions, which will lead to the accuracy of the analysis results decreased and the deviation increased. This paper proposes a combination filling method based on Maximum Likelihood Estimate (MLE), Multi-layer perceptron (MLP), and Light Gradient Boosting Machine (LightGBM) is proposed to realize different processing of multidimensional data. Taking the example of traffic accidents in the UK in 2022 as the data source, the MLE-MLP-LightGBM combined filling method proposed in this paper is compared with MLP, LightGBM and KNN. The experimental results show that the combination filling method is more accurate than the other three filling methods, and provides a more universal method for data filling.