Big Data for Insurance Companies

<p>Foreword xi<br />Jean–Charles POMEROL</p>
<p>Introduction &nbsp;xiii<br />Marine CORLOSQUET–HABART and Jacques JANSSEN</p>
<p>Chapter 1. Introduction to Big Data and Its Applications in Insurance 1<br />Romain BILLOT, C&eacute;cile BOTHOREL and Philippe LENCA</p>
<p>1.1. The explosion of data: a typical day in the 2010s 1</p>
<p>1.2. How is big data defined? &nbsp;4</p>
<p>1.3. Characterizing big data with the five Vs &nbsp;5</p>
<p>1.3.1. Variety &nbsp;6</p>
<p>1.3.2. Volume &nbsp;7</p>
<p>1.3.3. Velocity &nbsp;&nbsp;9</p>
<p>1.3.4. Towards the five Vs: veracity and value 9</p>
<p>1.3.5. Other possible Vs 11</p>
<p>1.4. Architecture 11</p>
<p>1.4.1. An increasingly complex technical ecosystem 12</p>
<p>1.4.2. Migration towards a data–oriented strategy 17</p>
<p>1.4.3. Is migration towards a big data architecture necessary? 18</p>
<p>1.5. Challenges and opportunities for the world of insurance &nbsp;20</p>
<p>1.6. Conclusion &nbsp;22</p>
<p>1.7. Bibliography &nbsp;&nbsp;23</p>
<p>Chapter 2. From Conventional Data Analysis Methods to Big Data Analytics 27<br />Gilbert SAPORTA</p>
<p>2.1. From data analysis to data mining: exploring and predicting &nbsp;27</p>
<p>2.2. Obsolete approaches 28</p>
<p>2.3. Understanding or predicting? 30</p>
<p>2.4. Validation of predictive models 30</p>
<p>2.4.1. Elements of learning theory 31</p>
<p>2.4.2. Cross–validation &nbsp;34</p>
<p>2.5. Combination of models 34</p>
<p>2.6. The high dimension case 36</p>
<p>2.6.1. Regularized regressions 36</p>
<p>2.6.2. Sparse methods &nbsp;38</p>
<p>2.7. The end of science? 39</p>
<p>2.8. Bibliography &nbsp;&nbsp;40</p>
<p>Chapter 3. Statistical Learning Methods &nbsp;43<br />Franck VERMET</p>
<p>3.1. Introduction &nbsp;43</p>
<p>3.1.1. Supervised learning 44</p>
<p>3.1.2. Unsupervised learning &nbsp;46</p>
<p>3.2. Decision trees &nbsp;46</p>
<p>3.3. Neural networks &nbsp;49</p>
<p>3.3.1. From real to formal neuron 50</p>
<p>3.3.2. Simple Perceptron as linear separator &nbsp;52</p>
<p>3.3.3. Multilayer Perceptron as a function approximation tool 54</p>
<p>3.3.4. The gradient backpropagation algorithm 56</p>
<p>3.4. Support vector machines (SVM) &nbsp;62</p>
<p>3.4.1. Linear separator &nbsp;62</p>
<p>3.4.2. Nonlinear separator 66</p>
<p>3.5. Model aggregation methods 66</p>
<p>3.5.1. Bagging &nbsp;67</p>
<p>3.5.2. Random forests &nbsp;69</p>
<p>3.5.3. Boosting &nbsp;&nbsp;70</p>
<p>3.5.4. Stacking &nbsp;&nbsp;74</p>
<p>3.6. Kohonen unsupervised classification algorithm &nbsp;74</p>
<p>3.6.1. Notations and definition of the model &nbsp;76</p>
<p>3.6.2. Kohonen algorithm 77</p>
<p>3.6.3. Applications &nbsp;79</p>
<p>3.7. Bibliography &nbsp;&nbsp;79</p>
<p>Chapter 4. Current Vision and Market Prospective 83<br />Florence PICARD</p>
<p>4.1. The insurance market: structured, regulated and long–term perspective 83</p>
<p>4.1.1. A highly regulated and controlled profession 84</p>
<p>4.1.2. A wide range of long–term activities &nbsp;85</p>
<p>4.1.3. A market related to economic activity &nbsp;87</p>
<p>4.1.4. Products that are contracts: a business based on the law &nbsp;87</p>
<p>4.1.5. An economic model based on data and actuarial expertise &nbsp;88</p>
<p>4.2. Big data context: new uses, new behaviors and new economic models 89</p>
<p>4.2.1. Impact of big data on insurance companies &nbsp;90</p>
<p>4.2.2. Big data and digital: a profound societal change 91</p>
<p>4.2.3. Client confidence in algorithms and technology 93</p>
<p>4.2.4. Some sort of negligence as regards the possible consequences of digital traces &nbsp;&nbsp;94</p>
<p>4.2.5. New economic models &nbsp;95</p>
<p>4.3. Opportunities: new methods, new offers, new insurable risks, new management tools &nbsp;95</p>
<p>4.3.1. New data processing methods &nbsp;96</p>
<p>4.3.2. Personalized marketing and refined prices 98</p>
<p>4.3.3. New offers based on new criteria &nbsp;100</p>
<p>4.3.4. New risks to be insured 101</p>
<p>4.3.5. New methods to better serve and manage clients 102</p>
<p>4.4. Risks weakening of the business: competition from new actors, uberization , contraction of market volume 103</p>
<p>4.4.1. The risk of demutualization 103</p>
<p>4.4.2. The risk of uberization 104</p>
<p>4.4.3. The risk of an omniscient Google in the dominant position due to data 105</p>
<p>4.4.4. The risk of competition with new companies created for a digital world 105</p>
<p>4.4.5. The risk of reduction in the scope of property insurance &nbsp;106</p>
<p>4.4.6. The risk of non–access to data or prohibition of use &nbsp;107</p>
<p>4.4.7. The risk of cyber attacks and the risk of non–compliance &nbsp;108</p>
<p>4.4.8. Risks of internal rigidities and training efforts to implement &nbsp;109</p>
<p>4.5. Ethical and trust issues 109</p>
<p>4.5.1. Ethical charter and labeling: proof of loyalty 110</p>
<p>4.5.2. Price, ethics and trust 112</p>
<p>4.6. Mobilization of insurers in view of big data &nbsp;113</p>
<p>4.6.1. A first–phase new converts &nbsp;113</p>
<p>4.6.2. A phase of appropriation and experimentation in different fields &nbsp;115</p>
<p>4.6.3. Changes in organization and management and major training efforts to be carried out &nbsp;118</p>
<p>4.6.4. A new form of insurance: connected insurance 118</p>
<p>4.6.5. Insurtech and collaborative economy press for innovation &nbsp;121</p>
<p>4.7. Strategy avenues for the future 122</p>
<p>4.7.1. Paradoxes and anticipation difficulties &nbsp;122</p>
<p>4.7.2. Several possible choices 123</p>
<p>4.7.3. Unavoidable developments 127</p>
<p>4.8. Bibliography &nbsp;&nbsp;128</p>
<p>Chapter 5. Using Big Data in Insurance &nbsp;131<br />Emmanuel BERTHEL&Eacute;</p>
<p>5.1. Insurance, an industry particularly suited to the development of big data 131</p>
<p>5.1.1. An industry that has developed through the use of data &nbsp;131</p>
<p>5.1.2. Link between data and insurable assets &nbsp;136</p>
<p>5.1.3. Multiplication of data sources of potential interest &nbsp;138</p>
<p>5.2. Examples of application in different insurance activities &nbsp;141</p>
<p>5.2.1. Use for pricing purposes and product offer orientation &nbsp;&nbsp;142</p>
<p>5.2.2. Automobile insurance and telematics &nbsp;143</p>
<p>5.2.3. Index–based insurance of weather–sensitive events 145</p>
<p>5.2.4. Orientation of savings in life insurance in a context of low interest rates 146</p>
<p>5.2.5. Fight against fraud 148</p>
<p>5.2.6. Asset management 150</p>
<p>5.2.7. Reinsurance &nbsp;150</p>
<p>5.3. New professions and evolution of induced organizations for insurance companies &nbsp;151</p>
<p>5.3.1. New professions related to data management, processing and valuation 151</p>
<p>5.3.2. Development of partnerships between insurers and third–party companies 153</p>
<p>5.4. Development constraints 153</p>
<p>5.4.1. Constraints specific to the insurance industry 153</p>
<p>5.4.2. Constraints non–specific to the insurance industry &nbsp;155</p>
<p>5.4.3. Constraints, according to the purposes, with regard to the types of algorithms used &nbsp;158</p>
<p>5.4.4. Scarcity of profiles and main differences with actuaries &nbsp;159</p>
<p>5.5. Bibliography &nbsp;&nbsp;161</p>
<p>List of Authors &nbsp;163</p>
<p>Index 165</p>