### Regression analysis and neural network based forecasting

Please suggest me that in Regression analysis and neural network based forecasting which is better? ..## Answers

It would actually depend on what you're trying to forecast. That is the important point.

For example, based on my limited knowledge, certain regression models require different assumptions to be met. (Normality of data, equal variances, etc.) They're great if you have a good idea of which independent variables affect the dependent variable(s).

Meanwhile, neural networks have been shown to be good at approximating chaotic systems/functions and predicting input patterns. ..

For example, based on my limited knowledge, certain regression models require different assumptions to be met. (Normality of data, equal variances, etc.) They're great if you have a good idea of which independent variables affect the dependent variable(s).

Meanwhile, neural networks have been shown to be good at approximating chaotic systems/functions and predicting input patterns. ..

It depends on the features and the regularization!

You might use the feature selection plugin of RapidMiner and test for each method, which feature set is better. Hence, it takes a set of processes: for each feature selection method and each learner you need to evaluate the performance. With RapidMiner it is easy, since the feature selection and the cross validation etc. is all readily available and you don't need to script all the processes. ..

You might use the feature selection plugin of RapidMiner and test for each method, which feature set is better. Hence, it takes a set of processes: for each feature selection method and each learner you need to evaluate the performance. With RapidMiner it is easy, since the feature selection and the cross validation etc. is all readily available and you don't need to script all the processes. ..

If your data follows iid normal distribution and does not have too many input attributes, regression analysis may be a good choice. Otherwise, ANN may be your choice.

Regression model was developed to find an estimated model such that error sum of square is smallest from all sample points to the model, and then infer it to population model. It is an overview model. However, ANN was developed to adjust the weighted coefficients between nodes in a multilayer structure by feeding each training example. It is more elaborated. ..

Regression model was developed to find an estimated model such that error sum of square is smallest from all sample points to the model, and then infer it to population model. It is an overview model. However, ANN was developed to adjust the weighted coefficients between nodes in a multilayer structure by feeding each training example. It is more elaborated. ..

Depends on with what accuracy and time efficiency u want....neural networkss are slow and has high processing time but accurate where as iid refers as two attributes not dependent on each other like date and time but time in hour min sec format can be dependent...... Go for neural first it's always a safe option but ur data must be a classification data