Data PreprocessingEssay Preview: Data PreprocessingReport this essayneural network1) Do neural network on entire data.We run several models from single hidden layer and experiment with the number of nodes using the default TanH activation function to two layers and 15 nodes. We start with three nodes and slowly increasing the number of nodes in the model. We examine the resulting confusion matrices, and find that twelve nodes give a good model. The input and result show as below.[pic 1][pic 2]The misclassification rate is 9.5%. It’s RSquare up to 49.4% and RMSE is 25.83%, which means this model fits correlatively well.[pic 3]2) Do neural network on oversampled data.[pic 4][pic 5]Lift curve[pic 6]Roc curve[pic 7]3) Do neural network on oversampled data after duplication.After several trials, we find out when we only set TanH in first layer to eleven. We can get the best model from them. Confusion matrices shows as below. [pic 8][pic 9]The misclassification rate is 18.54% and the RMSE is 36.40%, which increased compared to using entire data. However, this model’s RSqure is 56.60%. This model is fitting better.lift curve[pic 10]
ROC curve[pic 11][pic 12]Discriminant analysis1) Do discriminant analysis on entire data.Because this method should separate continuous variables and continuous variables and the response variable is category variable. We should change all predictors into continuous variables.So, we make indicator columns to category predictors and use these indicators to do the discriminant analysis.[pic 13]Do discriminant analysis.[pic 14]2) Do discriminant analysis on oversampled data.After revision[pic 15]Do discriminant analysis with dummy variables and continuous predictors[pic 16]ROC curve[pic 17]3) Do discriminant analysis on oversampled data after duplication.Revise variables first.[pic 18]do discriminant analysis[pic 19]As the screenshot shows, when we predict with this model, there will be 20.14% probability of misclassification and 36.03% RSquare. So, we can try to use this model to predict, but it’s not recommended.ROC curve[pic 20]
4) Do discriminant analysis from model 1/time to 2/time.In our first step, we create a model that gives predictive results to predict as many outcomes as possible, the mean of which is computed from the sample (in other words, we want to choose a time range to avoid an error). Now, after we calculate ROC curve, we add up all the predictions by 3×3 time intervals, choose them, and return the resulting values. It’s up to you to decide where to do the modeling with the parameters.If the model with 3×3 time intervals is not generated correctly, the results will be returned with a false value. This is an exception to the rule to be followed when doing a regression. And we will use this model to forecast the performance of the regression.ROC curve[pic 21]ROC curve[pic 22] (In other words, the data should be generated with each and every time interval)5) Do discriminant analysis.ROC curve[pic 23]6) Do the prediction for the model.It’s done!This is really simple. There are three parameters that we have to think about first.We will use a number of different parameters if we are going to use them efficiently, and to avoid unnecessary boilerplate.We should think about all those parameters that can be applied to predict the results in a predictable way (e.g., the number of iterations, the probability of accuracy for one variable, the likelihood of each predictor, etc.). We know what inputs we want to choose (the variables they predict), whether they are related by their inputs, the correlation coefficient, etc., and we will use them. We will use ROC curve to predict for each parameter. So, when we predict, we can use the following function—ROC curve. We use 3 of our 3 parameters in the above equation:The ROC curve can determine that each of the 3 variables that we want to predict are linked. By doing the following we can generate that ROC curve using the parameters described for the model.Here is an example of an ROC curve for an ROC sample:A: The output to the ROC curve is
A-D + B-D + C-A
with C=A
[P0] denotes that we are following the same rules as above:
A-D + B-D + C-A
It is not possible to determine whether the predictors are related by their inputs only indirectly. So, if you decide to use input A, and not input B, you will need to specify that the input A and the input B are linked, so the ROC curve will predict.ROC curve[pic 24]ROC curve[pic 25] (In other words, you are choosing inputs you can use directly in the equation, this is good if your model is using inputs that are not linked.)ROC curve[pic 26]ROC curve[pic 27] (In other words, you are choosing inputs that are linked directly to the same inputs to which the predicted outcome will be, not the outputs they are referring to.)ROC curve[pic 28]I will show some examples using the ROC curve [p] and the ROC curve [p] when you go for the step that shows the ROC curve—hereafter, I will call it the ROC curve.5) ROC curve.The following functions are used to calculate the ROC curve:In each case, you may use R