zoukankan      html  css  js  c++  java
  • Reliability diagrams

    Reliability diagrams (Hartmann et al. 2002) are simply graphs of the Observed frequency of an event plotted against the Forecast probability of an event. This effectively tells the user how often (as a percentage) a forecast probability actually occurred.

    In theory, a perfect forecast system will result in forecasts with a probability of X% being consistent with the eventual outcome X% of the time. Hence when plotting a reliability diagram comparisons are made against the diagonal (see below figure).

    With climate forecasting, it is also often the case that probabilities will be clustered together, generally near the 0.5 probability, as chaos dictates that situations are rarely perfectly predictable/historically identical, for similar initial conditions. Hence it is important to also plot some form of frequency distribution to indicate how the forecasts are spread over time. Are they all clustered together, or are they well spread? The below figure shows a probability function plotted for each curve, in order to demonstrate that the majority of forecasts were indeed clustered in the central probability bins.

    The technique for constructing the reliability diagram is similar to that for calculating the ROC score, only instead of plotting the hit rate against the false alarm rate, the hit rate is calculated only from the sets of forecasts for each probability separately. It is then plotted against the corresponding forecast probabilities.

    The hit rate for each probability bin is defined as:


    where O = number of Observed instances,
    NO = number of non-observed instances

    Frequency histograms are constructed from the same contingency tables as those used to produce the reliability diagrams. Frequency histograms show the frequency of forecasts as a function of the probability bin. The frequency of forecasts (Fn) for probability bin n is defined as:


    where T is the total number of forecasts.

    An example of a reliability diagram for Level 1 of the SVSLRF is given below:

    Reliability diagrams and frequency histograms are mandatory for level 1.

    References:

    Hartmann, H.C., Pagano, T.C., Sorooshiam, S. and Bales, R. 2002. Confidence builder: evaluating seasonal climate forecasts from user perspectives. Bull Amer. Met. Soc., 84, 683-698

  • 相关阅读:
    JUnit-执行测试
    JUnit注释的执行顺序
    回文数字
    JUnit简单测试
    利用jspx解决jsp后缀被限制拿shell
    wifi破解到局域网渗透
    爬虫之一:爬补天厂商数据(爬虫)
    三位一体的漏洞分析方法-web应用安全测试方法
    json.dumps错误:'utf8' codec can't decode byte解决方案
    nmap小技巧[1] 探测大网络空间中的存活主机
  • 原文地址:https://www.cnblogs.com/downtjs/p/3433021.html
Copyright © 2011-2022 走看看