基于用电特征单一视角数据的中小企业生命周期阶段识别

以我国某地区黑色金属加工及冶炼行业的212家中小企业为样本,先使用带行业影响修正的用电量增长率法标注样本,再利用K-means聚类算法和SMOTE过采样技术提取数据集特征和平衡训练集类别,最后采用经MetaCost元代价敏感算法改造后的梯度提升决策树模型进行企业生命周期阶段的识别建模和预测。构建的七项特征具有典型的长尾性质,表现在对模型的预测能力贡献上它们的重要度比较平均一致。经过代价矩阵调参,模型对失衡类别企业样本的最佳查准率和查全率分别为83.3%和88.9%。通过与传统方法结果的横向Kappa一致性检验和纵向实证分析,验证了基于单视角企业用电数据利用机器学习算法模型来识别企业生命周期阶段的可信性和有效性。

Life Cycle Phase Identification of Small and Medium-sized Enterprise Based on Single Perspective Data of Electricity

Like most lake wetlands located in the suburb, Chonghu Lake water system has some problems such as seasonal water shortage, poor water quality, single landform structure and lack of biological habitat. This paper starts with the study of the composition of material elements in the water environment, the overall spatial pattern, and the dynamic evolution rules of water system. Based on the basic principle of the hydrological process of the lacustrine wetland, this paper summarizes four influencing factors of the lake system that intervene in the hydrological process, such as water quantity, water quality, the shape of lake, and water biological habitat, and then puts forward a water environment restoration framework based on landscape approach. This paper, taking the Chonghu lake as an example, integrates landscape function and design method into the water environment restoration system according to the restoration scale. At the macro level, it is proposed to redistrict and construct regional water network and optimize the spatial morphology and structure of water system; at the meso level, it is proposed to dredge the surrounding water system and restore the balance of water quantity; at the micro level, it is proposed to build local water purification habitat and improve the habitat conditions. This paper is expected to provide references for research on conservation of water environment of lake wetlands in suburb and provide instructions for the development of the restoration work.