A fuzzy mining approach for energy efficiency in a Big Data framework
Abstract : The discovery and exploitation of hidden informa-tion in collected data has gained attention in many areas, particularly in the energy field due to their economic and environmental impact.Data mining techniques have then emerged as a suitable toolbox for analysing the data collected in modern network management systems in order to obtain a meaningful insight into consumption patterns and equipment operation. However, the enormous amount of data generated by sensors, occupational and meteorological data involve the use of new management systems and data processing. Big Data presents great opportunities for implementing new solutions to manage these massive data sets.In addition, these data present values whose nature complicates and hides the understanding and interpretation of the data and results. Therefore the use of fuzzy methods to adequately transform the data can improve their interpretability.This paper presents an automatic fuzzification method implemented using the Big Data paradigm, which enables, in a later step, the detection of interrelations and patterns among different sensors and weather data recovered from an office building
? Event detection is an important part in many Wireless Sensor Network (WSN) applications such as forest fire and environmental pollution.
? In this kind of applications,
? The event must be detected early in order to reduce the threats and damages.
? firstly, the results can vary a lot depending on the applied division, and secondly, this division may not be very intuitive for its later analysis of results.
? Fuzzy sets have been proven to adequately represent data with soft borders, increasing the interpretability of results by associating meaningful linguistic labels to the generated fuzzy sets.
? Other approaches use interval programming methods to tackle the uncertainty that the data may contain
In this paper we propose a fuzzification algorithm to ade-quately pre-process the data in order to apply, in a later step , fuzzy data mining techniques to discover potentially useful information that may be hidden in the data.
? we propose toautomatically fuzzify data collected by sensors and afterwardsapply fuzzy association rule mining to discover potentiallyuseful patterns in the field of energy management in buildings.
? To improve the interpretability of the datawe can modify the knowledge extraction algorithm throughthe use of fuzzy logic to create, for example, linguistic labelsfor sensors with numerical values that also provide meaningfulsemantics for the user.
|