1. API¶
- Summary
- Simply put,`kClusterLib` is a python tool to perform clustering of sensors using unprocessed/raw sensor data. It uses the famous pyCluster clustering library.
1.1. Introduction¶
For quick tryout of this work jump to usage document.
This library is composed of multiple modules kept together inside a single directory kClusterLib.
- Background
- Clustering is essentially an optimization process that tries to group similar sensors/objects based on degree of similarity in their(sensor’s) behavior/output( correlation ). The objective function( the parameter being minimized ) is the sum of intra cluster distances i.e. the measure of how close/similar are the element which are grouped together. However, the compactness of clusters(groups) is directly affected by the choice of number of clusters. This make the clustering a two dimentional problem. An optimum value for number of clusters is to be determined and at the same time grouping( clustering ) is to be performed.
The optimal clusters are recursively computed and are subject to
- Minimizing \(S_i = \frac{\displaystyle{\sum_{i=1}^{k}} \text{ (Intra Cluster Distances)}_i}{\displaystyle{\sum_{i=1}^{k}} \text{ (Inter Cluster Distances)}_i}\)
- Minimizing k: Number of clusters
1.2. Components¶
api : (current page) Contains brief information of the library work.bout the various parts and pieces of this work.
db connector: Database connection object and methods are defined in this module.
kc tools: Tools for data pre-processing are to be located here.
filterV: Filter transfer function is defined in this module.
usage: A demo example. | ...
