プレディクティブ・アナリティクスを使えるものに(英語版のみ)

  • Print
  • Connect
  • Email
  • Facebook
  • Twitter
  • LinkedIn
  • Google+
作成者 Michael Paczolt, William Torres |  2017年5月30日

プレディクティブ・アナリティクスとは何か、また何ではないか。

「プレディクティブ・アナリティクス」という用語は、ベンダーが専門用語から利益を得ようとして非常にゆがんだ意味になっているため、実質的に意味を成していません。これを明確に示すため、以下、インタラクティブな画像を使用します。

 

What predictive analytics IS NOT

 

Historical data summaries are not predictive analytics. While useful, these summaries, such as the one shown to the right, look backwards (retrospective) – akin to driving by looking in the rear view mirror. This summary of workers’ compensation claims data took less than 1 minute to create using basic data tools (e.g., a pivot table in Excel).

While historical data summaries can be useful to understand past experience, its weakness is just that, it’s retrospective. A data summary is descriptive analytics, not predictive analytics. Many vendors will try to convince you these are one and the same, but this is simply not true.

What predictive analytics IS

Predictive analytics is prospective – it looks forward. It aims to answer the question: “based on information available today, what is the predicted future outcome?” As you can imagine, this becomes much more complex. This is not as simple as summarizing data. Crystal balls are in short supply, so instead we rely on algorithms developed over the past few decades. An algorithm (i.e. a formula) is used to develop a model based on historical data. The model then returns a prediction for any new claim.

Imagine a new claim for a low back injury is reported for a 52-year-old worker with a history of prior injuries and comorbidities. Descriptive analytics would tell us that historically a low back claim has cost $20,000 on average. Predictive analytics would tell us that this claim is expected to cost $100,000 by considering all data related to the claim (such as worker’s age and medical history). Imagine how powerful this can be! Rather than being reactive, we can be proactive by preemptively identifying high-cost claims.

Parting thoughts

Despite the complexity of predictive analytics behind the scenes, an effective application of predictive analytics should be easy to understand for the end-user. The goal of predictive analytics is to make the end user’s work less complex (and save time).

The next time a vendor pitches you predictive analytics be skeptical. Determine if this information is retrospective or prospective. After all, predictive analytics is a truly powerful tool that will, and already has, revolutionized the way all businesses function.