ECNN: evaluating a cluster-neural network model for city innovation capability
ECNN: evaluating a cluster-neural network model for city innovation capability
复制标题
DOI：
10.1007/s00521-021-06471-z
复制
发表时间：
2021-09-18
影响因子：
6
通讯作者：
Wang, Xinyi
中科院分区：
文献类型：
--
作者：
Pei, Jiaming;Zhong, Kaiyang;Wang, Xinyi
创新能力是引领城市发展的强大动力。评价城市的创新能力对城市的发展也很重要。本文提出了一个评价城市创新能力的ECNN模型。该模型从机器学习的角度研究创新能力。与现有的统计方法相比，据我们所知，这是一个新的模型，可以从机器学习的角度来评估城市的创新能力。克服了原有统计方法研究指标间关系而不考虑指标与创新能力之间关系的不足。该模型首先对所有样本进行聚类，并将样本类别标记为聚类。其次，通过熵增益率计算各指标的权重，将各指标的权重值相加计算总分。为了获得更精确的结果，神经网络计算具有相同分数但属于聚类的样本分数，以良好的聚类数据作为训练集。这样，不同的集群就代表了不同的创新能力。每个样本都有一个创新能力得分。因此，ECNN模型在评价城市创新能力方面具有较高的实用性。
Innovation capability is a great driving force leading city development. It is also important to evaluate the innovation capability of a city for city development. In this paper, we propose an ECNN model to evaluate city innovation capability. This model studies innovation capability from the perspective of machine learning. Compared with the existing statistical methods, it is a novel model, to the best of our knowledge, to evaluate the city's innovation capability in terms of machine learning. It overcomes the shortcomings of the original statistical methods for studying the relationship between indicators without considering the relationship between indicators and innovation capabilities. This model first clusters all samples, and the sample categories are marked as clusters. Second, the weight of each indicator is calculated by the entropy gain rate, and the total score is calculated by adding the weighted values of each indicator. To obtain more precise results, the neural network calculates the sample scores, which have the same score but belong to the cluster, with good clustering data as the training set. In this way, different clusters represent different innovation capabilities. Each sample has an innovation capability score. Therefore, the ECNN model has high practicability in evaluating the innovation capability of cities.