Conception and Implementation of a Pattern Recognition System for Analysing Big Data from Distributed Embedded AMRA Boxes in Europe
There are several challenges in IoT Big Data analysis such as dealing with large amounts of missing input data and noisy and poor-quality data. These challenges lead to the inaccuracy of the classification methods when pattern recognition algorithms are applied on these raw data. The lack of pre-processing on raw data leads to an increase in complexity and miss-classification. In addition to these challenges, finding an appropriate classification method to deal with IoT big data challenges is essential.
Big Data Analysis can be considered as an important approach to extract useful information from a large amount of data which contains measured values from different sensors, device configuration and current software status. First of all, a pre-processing step should be performed on the given raw IoT data. The data representation and adaption process is one of the most important operations before feeding data into a classification method. After that, various classification approaches such as statistical and neural networks are considered to find the best-fit classification model for our Big Data problem.
The objective of this master thesis is to find and approach or a combination of approaches to solve the previously mentioned problem in the context of IoT Big Data Analysis based on pattern recognition algorithms. This particularly includes the state of the art regarding IoT Big Data analysis methods and classification approaches. The demonstration of feasibility with an implementation prototype of the concept using embedded AMRA boxes from Bosch is part of the thesis as well as a suitable evaluation including performance measures.