Classification

Search and Classification of "Interesting" Business Applications in the World Wide Web using a Neural Network Approach

Using NeuroSolutions, this study searched through a database of WWW businesses and classified them as “interesting” and “not interesting” for determining future business ventures. With generalized models, the new data sets were classified at 84.75% on average correctly.

Classification With Artificial Neural Networks and Supprt Vector Machines: Application to Oil Fluorescence Spectra

This paper reports on oil classification with fluorescence spectroscopy. Using NeuroSolutions, the objective is to classify the oil fluorescence spectra based on a laboratory dataset of fluorescence spectra of several oil classes (sludge, crude and heavy oil). The classification was carried out using the following three methods: channel relationship method (CRM), artificial neural networks (ANNs), and support vector machines (SVMs).

Neural Networks in Mobile Robot Motion

This paper deals with a path planning and intelligent control of an autonomous robot which should move safely in partially structured environment. This environment may involve any number of obstacles of arbitrary shape and size; some of them are allowed to move. We describe our approach to solving the motion-planning problem in mobile robot control using neural networks-based technique. Our method of the construction of a collision-free path for moving robot among obstacles is based on two neural networks. The first neural network is used to determine the “free” space using ultrasound range finder data. The second neural network “finds” a safe direction for the next robot section of the path in the workspace while avoiding the nearest obstacles. Simulation examples of generated path with proposed techniques will be presented

Using Satellite Imagery for Mapping Forest Types or Changes

Using remote sensing satellite imagery, multi-spectral sensor data is used for applications to map forest types or changes in a forest or to extract statistical information from the images together with available maps. In the case of mapping of forest types and changes, the information is essential for forestry companies in order to keep their databases up-to-date for planning and forest management. In addition, the imagery has been used to do wall-to-wall mapping of land use and vegetation from space as well as measuring the depth in the sea water, mapping along the coastlines and measuring the vegetation at the bottom of the sea. The primary benefit of using NeuroSolutions over other methods is not having to set up a mathematical model for the classification and regression applications. Common classification methods used in remote sensing is based on the assumption of normal distributed classes. However, in Mr. Rosengren's applications, the classes often have multimodal distributions. Using NeuroSolutions rather than other classification and regression methods, often save as much as 25% of the time and cost of the project!

Seabed Recognition Using Neural Networks

Side Scan Sonar (SSS) imaging is one of the advanced methods for data acquisition about the sea floor. The possibilities of intelligence-based approaches in the analysis of sonar images and classification of seabed material have been explored in this study. The only available type of measurement for classification SSS images is the grey level of the pixels corresponding to the acoustic reflectance. It is difficult to recognize and classify objects based on a single feature. However, the spatial order of the grey level transitions gives ‘texture’ characteristics to the image and it is these that act as an important aid in human interpretation. Image texture can be characterized by the Spatial Grey Level Dependence Method (SGLDM) based on the cooccurrence matrix of pairs of grey levels.

Snow Fall Prediction for the National Weather Service

Using NeuroSolutions, Mr. Roebber developed a snow density system that is currently employed by the National Weather Service to assist in the prediction of snow fall depths. Mr Roebber's system uses several different types of neural network architectures including the Principal Component Analysis (PCA) and Multilayer Perceptron (One and Two Hidden Layer) to complete his ensemble of 10 artificial neural networks. The ensemble correctly diagnoses 60.4% of the snow event cases examined, which is a substantial improvement over the 41.7% correct using the sample climatology, 45% correct using the 10-to-1 ratio (see table), and 51.7% correct using the National Weather Service "new snowfall to estimated meltwater conversion" table. The Heidke skill score measures the fraction of correct forecasts after eliminating those forecasts which would be correct due purely to random chance. The ensemble technique attains Heidke skill scores of 0.34 - 0.42, which is an increase of 75% - 183% over the next most skillful approach!

Corporate Financial Evaluation and Bankruptcy Prediction Implementing Artificial Intelligence Methods

Corporate accounting statements provide financial markets, and tax services with valuable data on the economic health of companies, although financial indices are only focused on a very limited part of the activity within the company. Useful tools in the field of processing extended financial and accounting data are the methods of Artificial Intelligence, aiming the efficient delivery of financial information to tax services, investors, and financial markets where lucrative portfolios can be created.

Predicting Sport Injuries and Player Performance Using Neural Networks

Mr. Murphy has integrated NeuroSolutions neural networks into several areas of professional sports including forecasting risk of injury, player performance and classifying match strategies. He uses approximately 20 to 30 different breadboards (including a mixture of classification and function approximation networks) such as: Multilayer Perceptron (MLP), Radial Basis Function (RBF), Modular Networks and Principal Component Analysis (PCA) networks for pre-processing. For predicting risk of injury the network occupies approximately 25 inputs including training loads over a 3-week period, wellness (i.e. fatigue, sleep quality & stress), pain & comfort ratings (i.e. foot, ankle, calf, groin, etc.) and player conditioning.

Neural Network Analysis of Interferometric Terahertz Images for Detection of Lethal Agents

A non-invasive means to detect and characterize concealed lethal agents employs spatial imaging of their characteristic transmission or reflection wavelength spectrum in the Terahertz (THz) electro-magnetic range. Artificial neural network (NN) analyses of these THz spectral images provide specificity of agent detection at reduced false alarm rates. Published THz spectra are utilized to generate simulated interferometric images of bioagent contained within an envelope, and a suicide bomber. Both multilayer perceptron and radial basis function NN architectures are used to analyze these spectral images. Positive identifications are generally made, with radial basis function NNs generally yielding superior results.

Pages

Subscribe to Classification