【作者】 杨威；

【导师】 胡亮；

【作者基本信息】 吉林大学 ， 软件工程， 2006， 硕士

【摘要】 中国网通(集团)公司鞍山市分公司从2005年开始,打破了从计费以来形成的十几年的传统处理模式,开始进行在月初开始进行每周进行计费带的市话处理工作,从2005年10月开始又陆续推广到所属县市分公司。实践证明,这些专题分析统计工作具有小、快、灵的特点,为企业经营管理提供信息支撑,进而强化业务发展质量和提升企业经济效益,提前把握市场运行的脉络,及时掌握市场动向。总之一句话,鞍山网通提高了竞争力。目前,这种提前分析预测的方式,从检索的资料中还没有看到,也没有现成的经验。在这种情况下,本文采用线形回归分析加逐渐逼近的方法,通过前几周的数据预测当月的通话量,并通过模型计算出当月的通话收入及当月总收入。经过两年的逐步完善,目前已经相对成熟,深受公司管理层的青睐。本文主要介绍了市话分析预测具体实现,包括范围,工作过程;利用描述统计和推断统计对每周的市话数据进行分析测算的过程;进行了可行性论证;介绍的测算收入模型;结合2个例子对每月测算产生差异进行了分析。更多还原

【Abstract】 Since 2005, the Anshan branch of China Netcom Group has begun to deal with the billing of local call every week, instead of the former traditional pattern. And after October, 2005, it was promoted to all towns in succession. It has been proved that this kind of analysis and statistic work is charactered with concise, rapid and efficient, which can provide necessary information to managers, reinforce the business development quality, improve the economic returns, grasp the track of market in advance, and get the market trend in the end. In a word, it help Anshan branch of CNC become more competitive.At present, there are not related theory and mature experience about this way of analysis and forecasting. In this case, this paper forecasts the call minutes according to the former weeks’data, using the linear regression analysis in addition to successive approximation. And then it builds a model to calculate the call revenue and the total revenue of the same month. After two years’practice and improvement, this model has been relatively mature, and well accepted by managers.This paper mainly introduces the implementation of local call’s analysis and statistics, including the scope and working process. The analysis of every week local call is based on the descriptive statistics and inferential statistics. And the research also includes feasibility demonstration, model’s building and the analysis of differences combining with two examples.The traditional billing begins at the end of each month, and the data collection, together with the handing in of revenue statements usually begins at 5th, next month. This paper gets a conclusion of this pattern after a primary exploration.The analysis of local call is carried on every week. Then the reckoning is based on the treated data, using the linear regression method. And the achievement of the call minutes prediction should be approached successively after the synthesis of several reckonings. The results prove that this method can basically meet our needs. And what’s more, it is feasible in theory.And another problem is to predict revenue. In this paper, model is built to solve this matter. Detailed introduce and three parameters are given in this paper.There must be some differences during the prediction of call minutes and revenue. The differences are discussed in theory and explained through practical examples. From the two years practices, the differences can be accepted to some extent, and meet the requirement of production operation.During the prediction of local call minutes, several influent phenomena are found. Some of them are direct, and easy to be understood. And some of them could be discovered only after thorough analysis. All of these phenomena take effect through some rules. Mangers could establish effective marketing strategies to improve our company’s competitive ability, instead of price war. It can get twice the result with half the effort. In this paper, an individual chapter is made to introduce the method.This prediction method which is charactered with concise, rapid and efficient can be used in many fields, such as Product Planning Centre, Marketing Department, VIP Centre and some like.At present, Anshan branch of CNC is carrying on the project of nation-wide intelligent solution. After this project, the Liaoning provincial company will achieve data online collection everyday. And then we could apply the analysis of revenue and call minutes in long distance call, phs, short messages and so on. There are more and more data, fields and unknowns waiting for us to explore.更多还原