Learning from Bad Data
نویسنده
چکیده
The data describing resolutions to telephone network local loop \troubles," from which we wish to learn rules for dispatching technicians , are notoriously unreliable. Anecdotes abound detailing reasons why a resolution entered by a technician would not be valid, ranging from sympathy to fear to ignorance to negligence to management pressure. In this paper, we describe four different approaches to dealing with the problem of \bad" data in order rst to determine whether machine learning has promise in this domain, and then to determine how well machine learning might perform. We then ooer evidence that machine learning can help to build a dispatching method that will perform better than the system currently in place.
منابع مشابه
The Rise of Patient Safety-II: Should We Give Up Hope on Safety-I and Extracting Value From Patient Safety Incidents?; Comment on “False Dawns and New Horizons in Patient Safety Research and Practice”
Who could disagree with the seemingly common-sense reasoning that: “We must learn from the things that go wrong.”? Despite major investments to improve patient safety, relatively few evaluations demonstrate convincing reductions in risk, harm, serious error or death. This disappointing trajectory of improvement from learning from errors or Safety-I as it is sometimes known has led some research...
متن کاملPrevalence of Bad Habits Among Juvenile Convicts
Objectives: This study was done to assess the spread of bad habits among juvenile convicts and determine the motives for use and the consequences of addiction. Methods: This study was descriptive and carried out at the Federal State Healthcare Institution. The study group included 106 adolescents aged 14 to 19 years. Sample selection was based on a multi-stage sampling method. The research too...
متن کاملFacilitating and Preventing Factors in Learning Clinical Skills from the Viewpoints of the Third Year Students of Fatemeh School of Nursing and Midwifery
Introduction: Investigating the problems and barriers in learning clinical skills has been regarded in so many studies but the factors facilitating this process have not been taken into consideration. This study was performed with the aim to determine the facilitating and preventing factors in learning clinical skills from the viewpoints of third year students of Fatemeh School of Nursing and M...
متن کاملFeature Selection in Big Data by Using the enhancement of Mahalanobis–Taguchi System; Case Study, Identifiying Bad Credit clients of a Private Bank of Islamic Republic of Iran
The Mahalanobis-Taguchi System (MTS) is a relatively new collection of methods proposed for diagnosis and forecasting using multivariate data. It consists of two main parts: Part 1, the selection of useful variables in order to reduce the complexity of multi-dimensional systems and part 2, diagnosis and prediction, which are used to predict the abnormal group according to the remaining us...
متن کاملCharacteristics of Data Suitable for Learning with Connectionist and Symbolic Methods
We contribute to a taxonomy of data with respect to appropriate selection of machine learning methods. Artificial data sets are carefully designed to expose the biases in a symbolic learning method and two connectionist learning methods. The data sets are based on the p-type and s-type classifications used in the past. Earlier results with these data sets are confirmed. New results indicate tha...
متن کامل