A useful approach to select relevant features from a dataset

A useful approach to select relevant features from a dataset is to use a random forest, an ensemble technique available in scikit-learn. Consider the Boston housing data set that gives housing values and other statistics in each of 506 suburbs of Boston based on a 1970 census. We used this data set before in the CIS 544 Data Mining and Machine Learning course, and explored random forest techniques using R. (You can your or file DecisionTrees.html uploaded to Ember). The random forest implementation in scikit-learn already collects feature importances for us so that we can access them via the feature_ importances_ attribute after fitting a RandomForestClassifier. Find the three most relevant features. The chunk of code below loads the Boston housing data set^2. import sklearn. datasets as sets data = sets. load_boston # Create a random forest model # from sklearn. ensemble import RandomForestClassifier #find the three most relevant features Read the first pages of Chapter 12 \"Training Artificial Neural Networks of our text- book (pages 341 to 347), and write a short summary of Modeling complex functions with artificial neural networks Single-layer neural network recap Use of IPython Notebook in Jupyter is recommended. Submissions should be in IPython notebook output in .pdf or hmtl, or both at the student\'s preference. Submissions of multiple files should be zipped up.

Solution

a) An artificial neural network is an artificial intelligence tool that sees and makes the production of arbitrary nonlinear multiparametric discriminant functions from the experimental data feeded in the system. The application of artificial neural network has attained an increased popularity for the applications where there was a use of mechanistic description of the dependency between dependent and independent variables. The application of artificial neural network to complex situations makes them highly usefule for the study of biological systems. The Biological systems being the most complex and unpredictable systems in terms of behaviour, the artifical intelligence recent applications consists the analysis of expression profiles ,genomic proteomic sequences and voice recognition.Therefore Artificial intelligence ability to model complex problems has resulted in their increased role in these tests.

b)The simplest kind of neural network is a single-layer perceptron network, that has single layer of output nodes. The inputs would be provided into the outputs with the series of weights.Like this,it could be considered the most basic kind of feed-forward network. The sum of the products of the weights and the inputs is processed for each node, and if the value shall be above a given specified threshold , the neuron would fire and would take up the activated value ; otherwise it incorporates the deactivated value.