The conventional aggregated performance measure (i.e., mean squared error) with respect to the whole dataset would not provide desired safety and quality assurance for each individual prediction made by a machine learning model in risk-sensitive regression problems. In this paper, we propose an informative indicator to quantify model reliability for individual prediction (MRIP) for the purpose of safeguarding the usage of machine learning (ML) models in mission-critical applications. Specifically, we define the reliability of a ML model with respect to its prediction on each individual input as the probability of the observed difference between the prediction of ML model and the actual observation falling within a small interval when the input varies within a small range subject to a preset distance constraint, namely , where denotes the observed target value for the input denotes the model prediction for the input , and is an input in the neighborhood of subject to the constraint . The developed MRIP indicator provides a direct, objective, quantitative, and general-purpose measure of “reliability” or the probability of success of the ML model for each individual prediction by fully exploiting the local information associated with the input and ML model. Next, to mitigate the intensive computational effort involved in MRIP estimation, we develop a two-stage ML-based framework to directly learn the relationship between and its MRIP , thus enabling to provide the reliability estimate for any unseen input instantly. Thirdly, we propose an information gain-based approach to help determine a threshold value pertaing to in support of decision makings on when to accept or abstain from counting on the ML model prediction. Comprehensive computational experiments and quantitative comparisons with existing methods on a broad range of real-world datasets reveal that the developed ML-based framework for MRIP estimation shows a robust performance in improving the reliability estimate of individual prediction, and the MRIP indicator thus provides an essential layer of safety net when adopting ML models in risk-sensitive environments.

中文翻译：

---

机器学习系统中个体预测的可靠性估计：基于模型可靠性的方法


相对于整个数据集的传统聚合性能测量（即均方误差）不会为机器学习模型在风险敏感回归问题中做出的每个单独预测提供所需的安全性和质量保证。在本文中，我们提出了一个信息指标来量化个体预测（MRIP）的模型可靠性，以保护机器学习（ML）模型在关键任务应用中的使用。具体来说，我们将 ML 模型相对于每个单独输入的预测的可靠性定义为，当输入在小范围内变化时，ML 模型的预测与实际观察之间的观察到的差异落入小区间内的概率到预设的距离约束，即 ， 其中 表示输入的观测目标值 表示输入 的模型预测，并且 是受约束 的邻域内的输入。开发的 MRIP 指标通过充分利用与输入和 ML 模型相关的本地信息，为每个单独的预测提供直接、客观、定量和通用的“可靠性”或 ML 模型成功概率的度量。接下来，为了减轻 MRIP 估计中涉及的密集计算工作，我们开发了一个基于 ML 的两阶段框架来直接学习 和 MRIP 之间的关系，从而能够立即为任何看不见的输入提供可靠性估计。第三，我们提出了一种基于信息增益的方法来帮助确定阈值，以支持何时接受或放弃依赖 ML 模型预测的决策。 在广泛的现实世界数据集上进行的全面计算实验以及与现有方法的定量比较表明，开发的基于机器学习的 MRIP 估计框架在提高个体预测的可靠性估计方面表现出强大的性能，因此 MRIP 指标提供了必要的在风险敏感环境中采用机器学习模型时的安全网层。