souravray97 / predictive-maintenance Goto Github PK
View Code? Open in Web Editor NEWThe Python and R code involves 3-4 techniques for predicting the classification of failures and non failures in the data. The following steps have been performed on the dataset: Data cleaning and/or one hot encoding for factor variables. Partitioning data into training and validation. Performing a logistic regression and predicting using the validation dataset. Lift and decile wise charts are constructed for the results obtained from the logistic regression performed. A classification tree has been built on the training dataset, the tree is pruned using the minimum cp value. A confusion matric for the tree has also been provided with an accuracy of 98.2%. A neural network with 1 hidden layer has been fit on the data.