Predicting Technical Problems of Hydropower Engineering Using eXtreme Gradient Boosting

Abstract

Nowadays, water shortage is increasingly severe, which has huge negative influence on daily life. Constructing hydropower engineering is one of the approaches to alleviate such problem. Therefore, it’s worth settling technical problems of hydropower engineering timely, which will help people not only make better use of water resources but also get rid of various security risks. To achieve such goal, this study predicts potential technical problems that hydropower engineering might happen. In order to utilize the large amount of data, data mining techniques are used to solve this multi-classification problem. First of all, plenty of data is preprocessed. Particularly, because of the complexity of text data, text mining techniques are applied to transform the unstructured data to structural data. Then, eXtreme Gradient Boosting (XGBoost) is applied to make the classification. To validate efficiency of the model, comparisons are made among XGBoost, Gradient Boosting Decision Tree, Random Forest, Decision Tree, k-Nearest Neighbor and Bernoulli Naive Bayes from the perspective of accuracy, precision, recall and f-score. The experimental result shows that XGBoost is more suitable to solve this classification problem. This study provides engineering inspectors with helpful suggestions of particular technical problems that need attention, and further enables people to inspect engineering more efficiently and effectively.