Predictive analytics: The power to predict who will click, buy, lie, or die
Predictive analytics: The power to predict who will click, buy, lie, or die
复制标题
预测分析：预测谁会点击、购买、撒谎或死亡的能力
DOI：
10.1057/jma.2013.14
复制
发表时间：
2013
影响因子：
3
通讯作者：
Richard Boire
中科院分区：
文献类型：
--
作者：
Richard Boire
为什么提前退休会降低预期寿命？为什么素食者错过的航班更少？这是两个更丰富多彩的例子，说明了在数据中等待的大量预测性发现。苏黎世大学发现，对于奥地利某个工作类别的男性来说，提前退休每增加一年，预期寿命就会减少1.8个月。他们推测这可能是由于退休后吸烟和饮酒等不健康的习惯造成的。一家航空公司发现，预订素食餐的顾客更有可能赶上航班，因为知道有个性化或特定的餐点等待顾客提供激励，或者建立一种承诺感。预测分析寻找这样的预测联系，然后研究它们如何联合收割机组合在一起，以实现更精确的预测。预测分析中最热门的趋势是什么？在预测分析的核心技术方面有许多令人兴奋的改进。一个是"隆起建模"（又名"隆起模型"）。“说服建模”），它预测影响力。. .来施加影响。奥巴马的竞选团队在2012年总统大选中利用它来影响选民;市场营销利用它来更熟练地说服客户;医学利用它来更好地选择每个病人的治疗方法。这个主题是本书最后一章的重点。另一个热门趋势是整体模特。就像集体智慧产生了一群人的智慧一样，我们在一群预测模型中看到了同样的效果。每个模型本身可能都相当原始，比如一些简单的规则，所以它的预测错误很多，就像一个试图预测的人一样。但是，如果他们作为一个群体聚集在一起，就会出现一个新的预测性能水平。内特银是否使用预测分析来预测奥巴马的选举？不--但奥巴马做到了。内特银对每个州的选举进行了整体预测：一个州的总体趋势是什么？与此同时，奥巴马的竞选团队正在使用预测分析来预测每个选民。除了预测之外，真正的力量来自于影响未来，而不是对未来进行猜测--这就是预测分析存在的理由。内特银公开竞争赢得选举预测，而奥巴马的分析团队则悄悄竞争赢得选举本身。具体来说，奥巴马团队通过对每张选票的预测来推动每张选民的竞选决策。预测分析所做的最酷的事情是什么？预测分析最具启发性的成就之一是IBM的沃森，它能够在电视智力竞赛节目Jeopardy!这些问题可以是关于几乎任何主题的，旨在供人类回答，并且可以是复杂的语法。事实证明，预测建模是沃森成功确定问题答案的方式：它预测，"这个候选答案是这个问题的正确答案吗？"它敲了一个又一个正确的答案-令人难以置信。公司对我作为客户的预测是什么？这里只是几个例子：微软帮助开发了一种技术，该技术基于GPS数据，可以提前多年准确预测一个人的位置。目标预测客户怀孕从购物行为，从而确定前景接触与新生儿的父母的需求相关的报价. Tesco（英国）每年在13个国家的杂货店收银机上发行1亿张个性化优惠券。预测分析将赎回率提高了3.6倍。Netflix赞助了一项价值100万美元的竞赛，以预测你会喜欢哪些电影，从而改善电影推荐。一家美国五大健康保险公司预测，老年保单保持器有可能在18个月内死亡，以触发临终咨询。爱迪生公司预测能源分配电缆故障，更新风险水平，显示在运营商的屏幕上每小时三次在纽约市。
Q & A with Author Eric Siegel Eric Siegel Why does early retirement decrease life expectancy and why do vegetarians miss fewer flights? These are two more colorful examples of the multitudes of predictive discoveries waiting within data. University of Zurich discovered that, for a certain working category of males in Austria, each additional year of early retirement decreases life expectancy by 1.8 months. They conjecture that this could be due to unhealthy habits such as smoking and drinking following retirement. One airline discovered that customers who preorder a vegetarian meal are more likely to make their flight, with the interpretation that knowledge of a personalized or specific meal awaiting the customer provides an incentive, or establishes a sense of commitment. Predictive analytics seeks out such predictive connections and then works to see how they may combine together for more precise prediction. What are the hottest trends in predictive analytics? There have been many exciting improvements in the core technology of predictive analytics. One is "uplift modeling" (a.k.a. "persuasion modeling"), which predicts influence . . . in order to do influence. The Obama campaign used it to influence voters in the 2012 presidential election; marketing uses it to more adeptly persuade customers; and medicine uses it to better select per-patient treatments. This topic is the focus of the final chapter of this book. Another hot trend is ensemble models. Like the collective intelligence that spawns the wisdom of a crowd of people, we see the same effect with a crowd of predictive models. Each model alone may be fairly primitive such as a few simple rules, so it gets prediction wrong a lot, as an individual person trying to predict also does. But have them come together as a group and there emerges a new level of predictive performance. Did Nate Silver use predictive analytics to forecast Obama's election? No--but Obama did. Nate Silver made election forecasts for each state as a whole: which way would a state trend, overall? In the meantime, the Obama campaign was using predictive analytics to make per-voter prediction. Moving beyond forecasting, true power comes in influencing the future rather than speculating on it--the raison d'tre of predictive analytics. Nate Silver publicly competed to win election forecasting, while Obama's analytics team quietly competed to win the election itself. Specifically, team Obama drove per-voter campaign decisions by way of per-vote predictions. What is the coolest thing predictive analytics has done? One of the most inspiration accomplishments of predictive analytics is IBM's Watson, which was able to compete against the all-time human champions on the TV quiz show Jeopardy! The questions can be about most any topic, are intended for humans to answer, and can be complex grammatically. It turns out that predictive modeling is the way in which Watson succeeds in determining the answer to a question: it predicts, "Is this candidate answer the correct answer to this question?" It knocks off one correct answer after another--incredible. What are companies predicting about me as a customer? Here are just a few examples: Microsoft helped develop technology that, based on GPS data, accurately predicts one's location up to multiple years beforehand. Target predicts customer pregnancy from shopping behavior, thus identifying prospects to contact with offers related to the needs of a newborn's parents. Tesco (UK) annually issues 100 million personalized coupons at grocery cash registers across 13 countries. Predictive analytics increased redemption rates by a factor of 3.6. Netflix sponsored a $1 million competition to predict which movies you will like in order to improve movie recommendations. One top-five U.S. health insurance company predicts the likelihood an elderly insurance policy holder will die within 18 months in order to trigger end-of-life counseling. Con Edison predicts energy distribution cable failure, updating risk levels that are displayed on operators' screens three times an hour in New York City.