【摘要】：隨著通信業(yè)的發(fā)展和人們生活水平提高,移動電信數(shù)據(jù)越來越反映著人們的生活狀況。為了研究移動電信數(shù)據(jù)與個人健康狀況之間的內(nèi)在關(guān)聯(lián),本文對移動電信數(shù)據(jù)進行跟蹤采集,并對移動電信數(shù)據(jù)進行分類提取特征值處理,同時進一步通過改進了極限學習機框架來預測電信用戶移動電信端的應用習慣和個人健康狀況之間的內(nèi)在聯(lián)系,從而對移動電信客戶進行健康風險預測和評估。本文以極限學習機理論為基礎(chǔ)進行改進,并使用改進后的算法作為本文的核心算法。改進后的算法作為一種強有力的分類器可通過調(diào)節(jié)系統(tǒng)參數(shù)并根據(jù)移動電信數(shù)據(jù)特征向量來預測客戶健康狀況。本文從移動電信數(shù)據(jù)中尋求和用戶自身相關(guān)的數(shù)據(jù)特征,進而從機器學習的角度出發(fā),基于極限學習機框架提出改進算法,對基于移動電信數(shù)據(jù)的個人健康狀況風險進行預測和評估,本文的研究致力對移動電信數(shù)據(jù)進行特征提取,同時對數(shù)據(jù)特征進行學習和分類識別。本文的主要工作包括:1.闡述了機器學習的基本理論思想及數(shù)學原理及三種數(shù)據(jù)挖掘算法的理論思想及數(shù)學原理,其中包含本文使用的極限學習機(ELM)框架,兩種對照算法為支持向量機(SVM)算法及反向傳播(BP)神經(jīng)網(wǎng)絡算法;2.針對移動電信數(shù)據(jù)的特點結(jié)合個人健康狀況相關(guān)指標,采用基于移動電信數(shù)據(jù)的采集、處理和數(shù)據(jù)整理的方法,將移動電信數(shù)據(jù)特點與個人健康相關(guān)主要指標進行關(guān)聯(lián)組建特征模型;3.提出了基于極限學習機框架的改進算法,用以處理移動電信數(shù)據(jù),對個人健康風險進行預測與評估。改進后的算法通過采用隱藏節(jié)點數(shù)子選擇過程,隨機選用隱藏節(jié)點向量來訓練網(wǎng)絡,通過選擇最佳的節(jié)點數(shù)來建立網(wǎng)絡參數(shù)和選擇訓練模型。同時,本文通過在不同的測試數(shù)據(jù)下和多個復雜的條件下的仿真實驗驗證了改進后的算法在此類場景下精確高效的數(shù)據(jù)處理能力,并與其他兩類算法進行了對比,證明改進后算法是一個高效準確且具有較低復雜度的數(shù)據(jù)識別算法。
[Abstract]:With the development of communication industry and the improvement of people's living standard, mobile telecommunication data more and more reflect people's living conditions. In order to study the inherent relationship between mobile telecommunication data and personal health, this paper tracks and collects mobile telecommunication data, and classifies and extracts the characteristic value of mobile telecommunication data. At the same time, the framework of extreme learning machine is improved to predict the inherent relationship between the application habits and personal health status of mobile telecommunication users, so as to predict and evaluate the health risk of mobile telecommunication customers. This paper is based on the theory of extreme learning machine, and uses the improved algorithm as the core algorithm of this paper. As a powerful classifier, the improved algorithm can predict the health status of customers by adjusting the system parameters and according to the feature vector of mobile telecommunication data. In this paper, from the point of view of machine learning, we propose an improved algorithm based on the framework of extreme learning machine, which is based on the mobile telecommunication data and the characteristics of the data related to the users themselves. The risk of personal health status based on mobile telecommunication data is predicted and evaluated. The research of this paper focuses on feature extraction of mobile telecommunication data, learning and classification of data features at the same time. The main work of this paper is as follows: 1. This paper expounds the basic theory and mathematical principle of machine learning and the theoretical and mathematical principles of three kinds of data mining algorithms, which includes the (ELM) framework of the extreme learning machine used in this paper. The two control algorithms are support vector machine (SVM) algorithm and backpropagation (BP) neural network algorithm. 2. According to the characteristics of mobile telecommunication data and the related indexes of personal health, this paper adopts the method of data collection, processing and data arrangement based on mobile telecommunication data. The characteristics of mobile telecommunication data are correlated with the main indicators of personal health to build a feature model. 3. An improved algorithm based on the framework of extreme learning machine is proposed to deal with mobile telecommunication data and to predict and evaluate personal health risks. The improved algorithm uses hidden node number sub-selection process, randomly selects hidden node vector to train the network, and sets up network parameters and training model by selecting the best number of nodes. At the same time, through the simulation experiments under different test data and multiple complex conditions, the improved algorithm is proved to be accurate and efficient in this kind of scenario, and is compared with the other two kinds of algorithms. It is proved that the improved algorithm is an efficient and accurate data recognition algorithm with low complexity.
【學位授予單位】：北京郵電大學
【學位級別】：碩士
【學位授予年份】：2017
【分類號】：R318;TN929.5

【相似文獻】

相關(guān)期刊論文 前5條

1 ;新美印尼合作發(fā)展亞洲移動電信網(wǎng)絡[J];特區(qū)科技;1994年05期

2 孫睿;蒲勇健;;移動電信低端增量市場中的價格策略互動[J];系統(tǒng)工程;2007年03期

3 余文斌;CDMA——中國移動電信市場的新亮點[J];廣東科技;2002年09期

4 安玉興;;資費模式與移動電信網(wǎng)絡寡斷競爭探討——基于LRT模型的分析[J];現(xiàn)代商貿(mào)工業(yè);2012年09期

5 ;Windows手機突破GPRS終端瓶頸[J];大眾科技;2002年09期

相關(guān)會議論文 前1條

1 車力軍;;移動電信運營商的業(yè)務創(chuàng)新研究[A];通信發(fā)展戰(zhàn)略與業(yè)務管理創(chuàng)新學術(shù)研討會論文集[C];2005年

相關(guān)重要報紙文章 前10條

1 王濤;澳門將增發(fā)移動電信服務臨時牌照[N];通信產(chǎn)業(yè)報;2000年

2 記者 成竹;浙江第三季度對移動電信類申訴同比增長一倍多[N];中國工商報;2009年

3 本報記者 葉勇;瑞銀：聯(lián)通將遭移動電信夾擊[N];上海證券報;2008年

4 ;德國3G移動許可證拍賣收槌[N];計算機世界;2000年

5 ;移動電信網(wǎng)絡管理系統(tǒng)[N];計算機世界;2001年

6 記者 蘆瑞;“被收費”“被服務”最惱人[N];河南日報;2010年

7 邵素宏;北京移動電信日啟動愛心助殘服務計劃[N];人民郵電;2008年

8 曾婭;十六家中企入圍移動電信榜上有名[N];人民郵電;2004年

9 吳彬 柯銘錕;黃石移動電信合作開新篇[N];人民郵電;2000年

10 廣東移動深圳分公司 熊明寶;手機媒體化：錢途無量[N];人民郵電;2008年

相關(guān)碩士學位論文 前3條

1 程瀟婷;基于移動電信數(shù)據(jù)的個人健康風險預測與評估[D];北京郵電大學;2017年

2 朱瑩;移動電信數(shù)據(jù)業(yè)務價值鏈評價方法及應用研究[D];東北大學;2008年

3 趙衛(wèi)華;基于支持向量機的移動電信行業(yè)客戶流失預測研究[D];昆明理工大學;2006年

，

本文編號：2441166