被装是军服、军鞋和军帽的统称，是展现国家军队形象和军人气质风貌的重要标志，也是衡量一个国家军队战斗力生成和战斗效能的重要方面。随着军队正规化、革命化、信息化建设程度不断提高，部队官兵对被装供应品种的选择性和舒适性提出了更高的要求。现行被装供应体系存在流程繁琐、保障人员多、保障对象涉及年龄、地域、民族、性别等差异较大、被装供应种类多数量大、带号型被装适体率不高等诸多难点问题，因此，将数据挖掘引入军装供应，通过对现有人员身形数据进行分析，发掘年龄、身高、体重、地区、民族等因素与人员身形数据的内在规律，对部队被装号型预测、物资预先储备、新兵军装供应适体率等方面有着重要的现实意义。论文主要工作如下：（1）对部队人员现有身形数据进行统计学分析。通过统计学方法分析一线军队个体的身高、体重、年龄、民族等身形特征数据，发现头围、脚长、身高、体重等身形特征数据的分布特点规律，其中身高、体重等身形数据不满足标准正态分布。利用灰色关联方法分析了年龄、地区、民族三个因素对身形数据的影响，发现三者与身形数据存在内在关联性，且因果关联性较大。进而利用层次分析法构建号型匹配模型，研究了身形数据与号型匹配之间的关系，提出了优化号型匹配的方案；（2）针对新兵身形数据测量晚于被装供应，且存在数据不全面、不准确和有缺失等实际问题，本文采用多种方法搭建神经网络模型进行适体性供应预测，利用年龄、地区和身高等三种数据分别建立军服和军鞋供应预测模型，并通过分析模型参数和训练过程来优化选择最佳的军服和军鞋预测模型；（3）利用实际数据进行模型应用和数据验证。采用本单位及外单位人员身形数据进行被装供应预测分析，结果表明均能提高10%以上的供应准确性，进一步验证了预测模型的准确性和可行性。本文开展被装供应号型辅助和号型预测模型的研究，能够有效纠正信息系统现有着装号型的错误，达到减少号型调换时间，减少保障成本，提高保障效率的目的。此外，研究成果能够为军队新品种研究预先储备提供决策依据，具有较强的实用价值。
Military uniforms are the general names of military clothes, military shoes and military hat, which is an important symbol to show the image of the national army and the style of military temperament. It is also an important aspect to measure the combat effectiveness of the national army. With the continuous development of our army‘s standardization, revolution and information construction, army officers and soldiers put forward higher requirements for the selectivity of military uniforms supply varieties and the comfort of wearing. There are many difficult problems in the current military uniforms supply system, such as cumbersome process, many supporting personnel, large differences in age, region, nationality and gender, many types and large quantities of uniforms supply, low fitness rate of numbered clothing, and so on. Therefore, data mining is introduced into uniforms supply, and the internal relationship of age, height, weight, region and nationality with personnel shape data are explored through the analysis of existing personnel shape data. It is of great practical significance for the prediction of military uniforms size, material pre-storage and fitness rate of recruits‘ uniforms supply. The main work of this thesis is as follows:(1) Statistical analysis on the existing body shape data of military personnel is carried out. By using statistical methods to analyze the body shape data of individual height, weight, age, nationality and other characteristics of the front-line army, the distribution characteristics of body shape characteristic data are found, such as head circumference, foot length, height and weight. It is also found that the characteristics such as height and weight do not meet the standard normal distribution. Then, by using the gray correlation method, the influence of age, region and nationality on body shape data is analyzed, and it is found that there is an internal correlation between them and body shape data, and there is a large causal correlation. Furthermore, using analytic hierarchy process, a clothing size matching model is constructed to study the relationship between body shape data and clothing size matching, and a scheme to optimize clothing size matching is proposed.(2) In view of the practical problems that the measurement of recruits‘ body shape data is later than the supply of clothing, and the data is incomplete, inaccurate and missing, many methods are used to build neural network models to predict the fitness supply of military uniforms. The prediction models of military uniforms (i.e. military clothes and military shoes) are established by using three kinds of available data of age, region and height. Then, by analyzing the model parameters of the methods, the best military clothes and military shoes models are optimized.(3) Some actual data are used for model application and data verification. The data of our unit and other units are used for prediction and analysis. It is found that the clothing supply accuracy can be improved by more than 10%, which further verifies the accuracy and feasibility of the prediction models.The above research results can be applied to the size assistance and size prediction of military uniforms supply, which can effectively correct the errors of the existing dress size in the information system, so as to reduce the size exchange time and the support cost and improve the support efficiency. In addition, the research results can provide decision-making basis for the research and pre-storage of new military varieties, and they have strong practical values.
