I-Corps: Creating data sets that are distilled into collections of information that may be used to answer questions in real-time

这个i-Corps项目的更广泛的影响/商业潜力是开发低成本、行业不可知的人工智能资源。目前，数据分析师和数据科学团队面临一个主要问题：数据整洁。数据清洁度是对记录集或表中损坏或不准确记录的度量。要清理数据库，必须识别数据的不完整或不正确部分，然后替换、修改或删除坏数据。机器学习模型通常在实验室条件下表现良好，但在现实世界中由于肮脏的数据而失败。由于噪声中的信令数据太少，人工智能解决方案在现实世界中的应用受到限制。这项拟议的技术解决了许多依赖手头肮脏数据的数据科学家面临的问题。几乎每个行业都需要更快、更好的决策。其目标是在完善内部流程、用户界面和初始技术目标金融机构的价值主张后，以可保护的知识产权和流程从战略上解决各个行业的数据清洁性问题。该i-Corps项目基于开发新型人工智能资源，以自动化数据准备、增强和治理过程。这些资源包括专有的机器学习算法、独特的数据库处理、动态训练集、下一代数据库技术和智能规则引擎，以重新设想昂贵、高投入和高度技术化的数据科学领域。该项目的新方法将自动化超过60%的数据科学家的工作量，降低成本，最大限度地减少内部IT资源需求，并增加信息访问。该奖项反映了NSF的法定使命，并通过使用基金会的智力优势和更广泛的影响审查标准进行评估，被认为值得支持。

The broader impact/commercial potential of this I-Corps project is the development of low-cost, industry agnostic, artificial intelligence resources. Presently, data analysts and data science teams face one primary problem: data cleanliness. Data cleanliness is a measure of corrupt or inaccurate records from a record set or table. To clean a database one must identify incomplete or incorrect parts of the data and then replace, modify, or delete the bad data. Machine learning models often perform well in lab conditions but fail in the real world due to dirty data. Artificial intelligence solutions are inhibited in real world applications due to too little signaling data amidst the noise. The proposed technology solves the problems that face many data scientists who dependent on the dirty data at hand. There is a demand for faster, better decision making across virtually every industry. The goal is to strategically solve the data cleanliness problem, with protectable intellectual property and processes, for various industries after refining internal processes, user interface, and value propositions for the initial technology target: financial institutions.This I-Corps project is based on the development of novel artificial intelligence resources to automate data preparation, augmentation, and governance processes. These resources include proprietary machine learning algorithms, unique data librarying processes, dynamic training sets, next generation database technology, and intelligent rules engines to reimagine the expensive, high-effort, and highly technical space of data science. The novel approach of this project will automate over 60% of a data scientists’ workload, reduce cost, minimize internal IT resource demands, and increase information access.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.