Alex J. Gutman - Becoming a Data Head

8 Chapter 10FIGURE 10.1 Fitting different logistic regression models to the data. The mo...FIGURE 10.2 Applying the logistic regression model to make predictions at GP...FIGURE 10.3 Simple decision tree applied to the HR intern datasetFIGURE 10.4 A random forest is a “forest” of several decision trees, usually...

9 Chapter 11FIGURE 11.1 A word cloud for the text in this chapterFIGURE 11.2 Processing text down to a bag of wordsFIGURE 11.3 Clustering documents and terms together with topic modeling. Can...

10 Chapter 12FIGURE 12.1 The simplest neural network possible. The four inputs are proces...FIGURE 12.2 A neural network with a hidden layer. The middle layer is “hidde...FIGURE 12.3 A deep neural network with two hidden layersFIGURE 12.4 Theoretical performance curves of traditional regression and cla...FIGURE 12.5 How a grayscale image “looks” to a computer, and how that data w...FIGURE 12.6 Color images are represented as 3D matrices for the pixel values...FIGURE 12.7 Convolution is like a series of magnifying glasses, detecting di...FIGURE 12.8 A simple representation of a recurrent neural networkFIGURE 12.9 Deep learning is a subfield of machine learning, which is a subf...

1 Cover Page

2 Table of Contents

3 Begin Reading

1 i

2 ii

3 v

4 vi

5 vii

6 ix

7 xi

8 xiii

9 xiv

10 xxiii

11 xxiv

12 xxv

13 xxvii

14 xxviii

15 xxix

16 xxx

17 xxxi

18 xxxii

19 xxxiii

20 xxxiv

21 xxxv

22 xxxvi

23 xxxvii

24 xxxviii

25 xxxix

26 xl

27 1

28 3

29 4

30 5

31 6

32 7

33 8

34 9

35 10

36 11

37 12

38 13

39 14

40 15

41 16

42 17

43 18

44 19

45 21

46 22

47 23

48 24

49 25

50 26

51 27

52 28

53 29

54 30

55 31

56 32

57 33

58 35

59 37

60 38

61 39

62 40

63 41

64 42

65 43

66 44

67 45

68 46

69 47

70 48

71 49

72 51

73 52

74 53

75 54

76 55

77 56

78 57

79 58

80 59

81 60

82 61

83 62

84 63

85 64

86 65

87 66

88 67

89 68

90 69

91 70

92 71

93 72

94 73

95 74

96 75

97 76

98 77

99 78

100 79

101 80

102 81

103 83

104 84

105 85

106 86

107 87

108 88

109 89

110 90

111 91

112 92

113 93

114 94

115 95

116 96

117 97

118 99

119 101

120 102

121 103

122 104

123 105

124 106

125 107

126 108

127 109

128 110

129 111

130 112

131 113

132 114

133 115

134 117

135 118

136 119

137 120

138 121

139 122

140 123

141 124

142 125

143 126

144 127

145 128

146 129

147 130

148 131

149 133

150 134

151 135

152 136

153 137

154 138

155 139

156 140

157 141

158 142

159 143

160 144

161 145

162 146

163 147

164 148

165 149

166 150

167 151

168 152

169 153

170 154

171 155

172 156

173 157

174 158

175 159

176 160

177 161

178 162

179 163

180 164

181 165

182 166

183 167

184 168

185 169

186 171

187 172

188 173

189 174

190 175

191 176

192 177

193 178

194 179

195 180

196 181

197 182

198 183

199 184

200 185

201 186

202 187

203 188

204 189

205 190

206 191

207 193

208 194

209 195

210 196

211 197

212 198

213 199

214 200

215 201

216 202

217 203

218 204

219 205

220 206

221 207

222 208

223 209

224 210

225 211

226 212

227 213

228 215

229 216

230 217

231 218

232 219

233 220

234 221

235 222

236 223

237 224

238 225

239 226

240 227

Becoming a Data Head is well-timed for the current state of data and analytics within organizations. Let's quickly review some recent history. A few leading companies have made effective use of data and analytics to guide their decisions and actions for several decades, starting in the 1970s. But most ignored this important resource, or left it hiding in back rooms with little visibility or importance.

But in the early to mid-2000s this situation began to change, and companies began to get excited about the potential for data and analytics to transform their business situations. By the early 2010s, the excitement began to shift toward “big data,” which originally came from Internet companies but began to pop up across sophisticated economies. To deal with the increased volume and complexity of data, the “data scientist” role arose with companies—again, first in Silicon Valley, but then everywhere.

However, just as firms were beginning to adjust to big data, the emphasis shifted again—around about 2015 to 2018 in many firms—to a renewed focus on artificial intelligence. Collecting, storing, and analyzing big data gave way to machine learning, natural language processing, and automation.

Embedded within these rapid shifts in focus were a series of assumptions about data and analytics within organizations. I am happy to say that Becoming a Data Head violates many of them, and it's about time. As many who work with or closely observe these trends are beginning to admit, we have headed in some unproductive directions based on these assumptions. For the rest of this foreword, then, I'll describe five interrelated assumptions and how the ideas in this book justifiably run counter to them.

Assumption 1: Analytics, big data, and AI are wholly different phenomena.

It is assumed by many onlookers that “traditional” analytics, big data, and AI are separate and different phenomena. Becoming a Data Head, however, correctly adopts the view that they are highly interrelated. All of them involve statistical thinking. Traditional analytics approaches like regression analysis are used in all three, as are data visualization techniques. Predictive analytics is basically the same thing as supervised machine learning. And most techniques for data analysis work on any size of dataset. In short, a good Data Head can work effectively across all three, and spending a lot of time focusing on the differences among them isn't terribly productive.

Assumption 2: Data scientists are the only people who can play in this sandbox.

We have lionized data scientists and have often made the assumption that they are the only people who can work effectively with data and analytics. However, there is a nascent but important move toward the democratization of these ideas; increasing numbers of organizations are empowering “citizen data scientists.” Automated machine learning tools make it easier to create models that do an excellent job of predicting. There is still a need, of course, for professional data scientists to develop new algorithms and check the work of the citizens who do complex analysis. But organizations that democratize analytics and data science—putting their “amateur” Data Heads to work—can greatly increase their overall use of these important capabilities.

Assumption 3: Data scientists are “unicorns” who have all the skills needed for these activities.