Deep online hierarchical dynamic unsupervised learning for pattern mining from utility usage data

Abstract

While most non-intrusive load monitoring (NILM) work has focused on supervised algorithms, unsupervised approaches can be more interesting and practical. Specifically, they do not require labelled training data to be acquired from the individual appliances and can be deployed to operate on the measured aggregate data directly. We propose a fully unsupervised novel NILM framework based on Dynamic Bayesian hierarchical mixture model and Deep Belief network (DBN). The deep network learns, in unsupervised fashion, low-level generic appliance-specific features from the raw signals of the house utilities usage, then the hierarchical Bayesian model learns high-level features representing the consumption patterns of the residents captured by the correlations among the low-level features. The temporal ordering of the high-level features is captured by the Dynamic Bayesian Model. Using this architecture, we overcome the computational complexity that would occur if temporal modelling was directly applied to the raw data or even to the constructed features. The computational efficiency is crucial as our application involves massive data from different utilities usage. Moreover, we develop a novel online inference algorithm to cope with this big data. Finally, we propose different evaluation methods to analyse the results which show that our algorithm finds useful patterns.

User Behavior Traffic Analysis Using a Simplified Memory-Prediction Framework

“…Furthermore, as more experiments were conducted, it was found that grid cells, such as neurons, may also be present in the neocortex [21,22]. The memory-prediction framework has been implemented in various fields, including object identification [23,24], medicine [25], online learning [26,27], real-time network traffic anomaly detection [28][29][30][31], and network forensics [32].…”

Renewable quantile regression for streaming data sets

“…Applying statistical models and methods to such big data can cause excessive computational burden not only in terms of strains on computer memory due to the large volume but also strains in terms of computational efficiency since even seemingly very simple tasks can take an inordinate amount of time to compute. In a recent review, [31] grouped the statistical and computational methodologies into three categories: subsampling-based approaches (e.g., [35,40; 34]), divide and conquer approaches (e.g., [23]; [18]; [1]; [16]; [17]), and online updating approaches (e.g., [27]; [24]; [28]; [26]). Online updating approaches are distinct from the other two because they target problem where data arrive in streams or large chunks and address statistical problems in an updating framework without storage requirements for previous data.…”