移動通信大數據分析：數據挖掘與機器學習實戰(zhàn)

第1章概述
1.1 電信業(yè)大數據分析 ···························1
1.2 電信大數據分析的驅動力 ················2
1.3 大數據分析對電信產業(yè)價值鏈的
益處 ··················································3
1.4 電信大數據的實現(xiàn)范圍····················4
1.4.1 網絡分析 ···················································5
1.4.2 用戶與市場分析 ·······································8
1.4.3 創(chuàng)新的商業(yè)模式 ·······································91.5 本書概要 ··········································9
參考文獻 ·················································10
第2章電信分析方法論
2.1 回歸方法 ········································12
2.1.1 線性回歸 ··················································13
2.1.2 非線性回歸 ··············································15
2.1.3 特征選擇 ··················································16
2.2 分類方法 ········································18
2.2.1 邏輯回歸 ··················································18
2.2.2 其他分類方法 ··········································19
2.3 聚類方法 ········································20
2.3.1 K均值聚類 ··············································21
2.3.2 高斯混合模型 ··········································23
2.3.3 其他聚類方法 ··········································24
2.3.4 聚類方法在電信數據中的應用 ·················25
2.4 預測方法 ········································25
2.4.1 時間序列分解 ··········································26
2.4.2 指數平滑模型 ··········································27
2.4.3 ARIMA模型 ············································28
2.5 神經網絡和深度學習 ·····················29
2.5.1 神經網絡 ··················································29
2.5.2 深度學習 ··················································31
2.6 強化學習 ········································32
2.6.1 模型和策略 ··············································33
2.6.2 強化學習算法 ··········································33
參考文獻 ·················································34

XII
XII

第3章 LTE網絡性能趨勢分析
3.1 網絡性能預測策略 ·························39
3.1.1 直接預測策略 ··········································39
3.1.2 分析模型 ··················································39
3.2 網絡資源與性能指標之間的關系 ···40
3.2.1 LTE網絡KPI與資源之間的關系 ···········40
3.2.2 回歸模型 ··················································41
3.3 網絡資源預測 ·································43
3.3.1 LTE網絡流量與資源預測模型 ···············43
3.3.2 預測網絡資源 ··········································43
3.4 評估RRC連接建立的應用 ············46
3.4.1 數據準備與特征選取 ······························46
3.4.2 LTE KPI與網絡資源之間的關系推導 ····47
3.4.3 預測RRC連接建立成功率 ·····················49
參考文獻 ·················································50
第4章熱門設備就緒和返修率分析
4.1 設備返修率與設備就緒的預測
策略 ················································53
4.2 設備返修率和就緒預測模型 ··········54
4.2.1 預測模型的移動通信服務 ························54
4.2.2 參數獲取與存儲 ······································55
4.2.3 分析引擎 ··················································56
4.3 實現(xiàn)和結果 ·····································58
4.3.1 設備返修率預測 ······································58
4.3.2 設備就緒預測 ··········································62
第5章 VoLTE語音質量評估
5.1 應用POLQA評估語音質量··········68
5.1.1 POLQA標準···········································68
5.1.2 語音質量評價中的可擴展性和
可診斷性 ··················································69
5.2 CrowdMi方法論 ····························69
5.2.1 基于RF特征的分類 ·······························70
5.2.2 網絡指標選擇與聚類 ······························70
5.2.3 網絡指標與POLQA評分之間的關系····70
5.2.4 模型測試 ··················································70
5.3 CrowdMi中的技術細節(jié) ·················71
5.3.1 記錄分類 ··················································71
5.3.2 網絡指標的選擇 ······································71
5.3.3 聚類 ·························································72
5.3.4 回歸 ·························································73
5.4 CrowdMi原型設計與試驗 ·············74
5.4.1 客戶端和服務器架構 ······························74
5.4.2 測試和結果 ··············································76
參考文獻 ·················································78

目 錄XIII

目 錄XIII
第6章移動APP無線資源使用分析
6.1 起因和系統(tǒng)概述 ·····························80
6.1.1 背景和挑戰(zhàn) ··············································80
6.1.2 移動資源管理 ··········································81
6.1.3 系統(tǒng)概述 ··················································82
6.2 AppWiR眾包工具 ··························83
6.3 AppWiR挖掘算法 ··························84
6.3.1 網絡指標的選擇 ······································84
6.3.2 LOESS方法 ············································87
6.3.3 基于時間序列的網絡資源使用預測 ·······87
6.4 實現(xiàn)和試驗 ·····································88
6.4.1 數據收集與研究 ······································88
6.4.2 結果和準確度 ··········································89
參考文獻 ·················································91
第7章電信數據的異常檢測
7.1 模型 ················································93
7.1.1 高斯模型 ··················································94
7.1.2 時間依賴的高斯模型 ······························94
7.1.3 高斯混合模型（GMM）·························95
7.1.4 時間依賴的高斯混合模型 ·······················95
7.1.5 高斯概率潛在語義模型（GPLSA）·······95
7.2 模型對比 ········································97
7.2.1 樣本定義 ··················································97
7.2.2 異常識別 ··················································98
7.2.3 時間依賴GMM與GPLSA的對比 ·········997.3 仿真與討論 ···································100
參考文獻 ···············································103
第8章基于大數據分析的LTE網絡自優(yōu)化
8.1 SON（自組織網絡）···················105
8.2 APP-SON ······································107
8.3 APP-SON架構 ·····························108
8.4 APP-SON算法 ·····························110
8.4.1 匈牙利算法輔助聚類（HAAC）··········111
8.4.2 單位回歸輔助聚類數的確定 ·················114
8.4.3 基于DNN的回歸·································114
8.4.4 每個小區(qū)在時序空間的標簽組合 ·········116
8.4.5 基于相似性的參數調整 ·························1168.5 仿真與討論 ···································117
參考文獻 ···············································122
第9章電信數據和市場營銷
9.2.1 數據采集和數據類型 ····························130
9.1 電信營銷專題 ·······························127
9.2.2 網絡的提取和管理 ································131
9.2 社交網絡的總體構建 ···················130



9.3 網絡結構的度量 ···························133
參考文獻 ···············································135

9.4 網絡中的消費者行為建模 ············134
第10章傳染式客戶流失
10.1 問題引入 ·····································138
10.1.1 流失率問題 ··········································138
10.1.2 社交學習和網絡效應 ··························139
10.2 網絡數據的處理 ·························141
10.3 動態(tài)模型 ·····································143
10.3.1 模型介紹 ··············································143
10.3.2 模型的定義 ··········································144
10.3.3 自身經驗建模、社交學習和
社交網絡效應 ······································146
10.3.4 模型估計 ··············································148
10.4 結果 ············································149
參考文獻 ···············································151
第11章基于社交網絡的精準營銷
11.1 網絡效應的渠道 ·························158
11.2 社交網絡數據處理 ·····················159
11.3 建模策略問題 ·····························160
11.3.1 線性空間自回歸模式 ···························160
11.3.2 社交網絡交互模型 ······························162
11.3.3 內生同伴效應 ······································162
11.4 發(fā)現(xiàn)與應用 ·································164
11.4.1 結果的解釋 ··········································164
11.4.2 基于社交網絡的精準營銷 ···················165
參考文獻 ···············································168
第12章社交影響和動態(tài)社交網絡結構
12.1 動態(tài)模型 ·····································17712.1.1 連續(xù)時間馬爾可夫模型假設 ···············17712.1.2 模型估計與識別 ··································17912.1.3 網絡結構對社交影響的多元分析 ·······18012.2 研究發(fā)現(xiàn)總結 ·····························18112.2.1 隨機行動者動態(tài)網絡模型的
估計結果··············································182

12.2.2 元回歸分析結果 ··································184
12.2.3 策略模擬 ··············································18812.3 結論 ············································193
參考文獻 ···············································194