首先是分析大資料 (big data) (註 1)：
Here’s how much data Friedberg and his team of 21 Ph.Ds. have and how they analyze it:
• About 200TB of historical data stored in Amazon S3.
• Monthly predictive simulations and daily forecasts for each location it insures.
• Each simulation:
- Uses between 4,000 and 5,000 computing cores on Amazon Elastic MapReduce.
- Considers 10,000 scenarios for each of the next 730 days.-- Analyzes 5 trillion data points.
- Consumes 20TB of uncompressed data.
• Data volumes have grown about 15x since last year.
然後，每個農夫根據其不同的需求，和 Climate Corp. 簽不同的合約。當天氣影響到農作物的收成時，農夫得到理賠。相對於政府的繁瑣手續 (註 2)：
Historically, farmers have relied on crop insurance sold by the federal government. The program is rife with all the red tape bureaucracies are famous for. Farmers must plant their fields on a schedule determined by the government and consent to inspections. To estimate crop value, they have to record and turn over years’ worth of data about yields. When disaster hits, claims take months to process, and the payout often only covers costs, not lost profits.Climate Corp. 提供迅速地自動理賠 (註 2)：
Climate Corp. approaches crop insurance with the typical Silicon Valley emphasis on speed and automation. Farmers log onto the startup’s site, input a few details about their planting plans, and an algorithm dubbed the Farm-Level Optimizer spits back policy options. Because its servers are always recording weather data, Climate Corp. knows when a policy is triggered by, say, a drought and automatically issues payments to the affected farmers. Climate Corp. also opens up its data trove to its customers, showing them the range of yields and profits they can expect based on likely weather conditions. Friedberg says thousands of farmers have insured millions of acres through the startup, though the company won’t divulge revenue details.(註 1) Derrick Harris, How Climate Corp. is pitting big data against Mother Nature, May 2, 2012.
(註 2) Ashlee Vance, Climate Corp. Updates Crop Insurance via High Tech, March 22, 2012
沒有留言:
張貼留言