摘要
由一名新闻记者-机器人与机器学习的工作人员新闻编辑每日新闻-调查人员发布了关于马学习的新报告。根据NewsRx编辑在印度Uttarakhand的新闻报道，研究表明：“滑坡是印度喜马拉雅地区Beas河谷上游最具破坏性的灾害之一，为了预测该地区滑坡易发区的空间变异性，绘制滑坡易感性图是一项重要的初步工作。”这项研究的财政支持来自印度科学和工业研究理事会(CSIR)。我们的新闻记者从Wadia Institute of Mayrama Geologies获得了一句研究的引文，“由于机器学习算法的使用增加了敏感性研究的成功率，四种机器学习模型的性能，即Naive Bayes(NB)、K-最近邻居(KNN)、对随机Fores T(RF)和极端梯度提升(XGBoost)进行了初步的滑坡敏感性制图试验，利用包含滑坡和非滑坡数据的滑坡调查和13个滑坡条件因子对模型进行了训练，采用超参数优化和基于变重要度的输入因子选择对模型进行了优化，其中极端梯度提升(XGBoost)、基于集成的先进Machin E学习算法表现出优异的性能（AUC=相似于0.91），其次是RF、NB和KNN，AUC值相似于0.88、相似于0.87和相似于0.82.。因此，选择XGboost模型进行详细研究。结果表明,滑坡易感区占总面积的44%,位于公路和水系附近,坡度在31°～50°的南向边坡中,滑坡易感区主要分布在公路和水系附近,坡度在31°～50°之间。
Abstract
By a News Reporter-Staff News Editor at Robotics & Machine Learning Daily News Daily News-Investigators publish new report on Ma chine Learning. According to news reporting out of Uttarakhand, India, by NewsRx editors, research stated, "Landslide is one of the most destructive hazards in the Upper Beas valley of the Himalayan region of India. Landslide susceptibility mapping is an important and preliminary task in order to prospect the spatial v ariability of landslide prone zones in the area." Financial support for this research came from Council of Scientific & Industrial Research (CSIR) - India. Our news journalists obtained a quote from the research from the Wadia Institute of Himalayan Geology, "As the use of machine learning algorithms has increased the success rate in susceptibility studies, the performance of the four machine learning models, namely Naive Bayes (NB), K-Nearest Neighbor (KNN), Random Fores t (RF) and Extreme Gradient Boosting (XGBoost) were initially tested for landsli de susceptibility mapping in the area. Landslide inventory containing both lands lide and non-landslide data and thirteen landslide conditioning factors were con sidered to train the models. The models were optimized using hyperparameter opti mization and input factors selection based on variable importance. Among the fou r models, Extreme Gradient Boosting (XGBoost), an advanced ensemble-based machin e learning algorithm, demonstrated superior performance (AUC = similar to 0.91) followed by RF, NB and KNN with AUC values of similar to 0.88, similar to 0.87, and similar to 0.82. Therefore, XGboost model was selected for detailed study, i ncluding sensitivity analysis. The results depict that 44% of the total area falls under high and very high susceptible zones. Southward facing sl opes having inclination between 31 degrees-50 degrees located at an elevation of 2001-3000 m in the vicinity of road and drainage network contain most of the la ndslide susceptible zones."
NETL
NSTL