Research on Overseas Automotive Data Processing Based on Feature Engineering and Machine Learning

In the context of the globalization of the automotive industry, text data such as user reviews, policy documents, and maintenance records in overseas automotive data contain rich information, but they are difficult to process due to issues such as language diversity and semantic complexity. This paper first acquires relevant multi-source heterogeneous data based on distributed crawler technology, then uses feature engineering methods to perform data cleaning, outliers and missing values on numerical data, and uses data transformation to perform word segmentation, word vector representation, etc. Then, a new multisource heterogeneous dataset is constructed by linking numerical and text data using splicing techniques; Finally, XGBOOST technology is used to optimize regional sales volume, export logistics cost and market risk warning for the processed multi-source heterogeneous data set to achieve high precision prediction. The aim is to provide innovative ideas and practical paths for overseas automotive data processing and help automotive companies enhance their competitiveness in overseas markets.