This project was a part of my tasks during my internship at Stir Research. The project consisted of various data exploratory analysis and data visualizations to be carried out on the given data by the company.
Language: Python
Modules and Packages: numpy, matplotlib, pandas, scikit-learn, seaborn, keras, os, csv, time
Dataset: Corrosion Resistance dataset by the Stir Research
This is a Data Analytics project with the following tasks:
-
Clean the dataset and Create Visualizations of each of the properties of welding metal against Corrosion pit potential/mv
-
Create an Artificial Neural Network model to predict Corrosion resistance.
-
Plot the Mean Squared error and Mean Absolute Percentage error during training.
-
Visualize a 3D Surface Plot and 2D Contour Plot for Corrosion resistance, Rotational Speed, Welding Speed and Axial Force.
-
Visualize and Tri Surface Plot for the N, S and F.
Task 1: Clean the dataset and Create Visualizations of each of the properties of welding metal against Corrosion pit potential/mv
Step 1: Import required libraries
Step 2: Import and read the Dataset
Step 3: Clean the dataset using Ms Excel
Step 4: Visualize a bar plot using seaborn of corrosion resistance vs corrosion pit potential/mv
Step 5: Use seaborn again to viuzalie bar plot by tool profile.
Step 6: Do the same for Rotartional Speed, Welding Speed, and Axial Force.
Step 7: Visualize a heatmap for all the properties of the tools and a bar plot for the no. of experiments conducted
Task 2: Create an Artificial Neural Network model to predict Corrosion resistance.
Step 1: Import required libraries
Step 2: Import and read the Dataset
Step 3: Split datasets in input values(x) and output values(y), data pre-processing for regression
Step 4: Standardize dataset and reshape the dataset
Step 5: Split into train and test
Step 6: Nh=Ns/(Ξ±β(Ni+No)) ie; 30/(2(3+1)) = 3.75 approx 4
- the first hidden layer has 4 nodes and uses relu
- the second hidden layer has 3 nodes and uses relu activation function
- the output layer had one node and uses sigmoid activation function
- setting input layer with right no. of input features by setting it to 3 for 3 input features
Step 7: Model will be fit with stochastic gradient descent with a learning rate of 0.01 and momentum of 0.9.
Step 8: Training will be performed for 100 epochs and test set will be evaluated at the end of each epoch
Step 9: Evaluating the model/ training and testing for mean squared error
Step 10: Plot mean squared error during training.
Results
Task 3: Plot the Mean Squared error and Mean Absolute Percentage error during training.
Step 1-Step 10 of Task 2 repeated and then follow the same 10th step mean Absolute percentage error.
Results
Task 4: Visualize a 3D Surface Plot and 2D Contour Plot for Corrosion resistance, Rotational Speed, Welding Speed and Axial Force.
Step 1: Import required libraries
Step 2: Import and read the Dataset
Step 3: Define the Co-ordinates by meshgrid
Step 4: Set the plot as 3D projection and define it as contour for corrosion resistance.
Step 5: Repeat from Step 3 and 4 for rotational speed, welding speed and Axial force contours.
Results
Task 5: Visualize and Tri Surface Plot for the N, S and F.
Step 1: Import required libraries
Step 2: Import and read the Dataset
Step 3: Transform it to a long format And transform the old column name in something numeric
Step 4: Define the plot as trisurface
Step 5: Set the labels, title and plot the tri surface and Add a color bar which maps values to colors.
Results
