s-b-iqbal/steel-plates-fault-diagnosis-using-classification-models

The objective of the project is to classify steel plates fault into 7 different types. The end goal is to train several machine Learning Algorithms for automatic pattern recognition.

Steel Plates fault diagnosis using Classification Models

Objective:

The objective of the project is to classify steel plates fault into 7 different types. The end goal is to train several machine Learning Algorithms for automatic pattern recognition. In doing so, I have employed Logistic Regression, KNN, Naïve Bayes, Decision Trees and Random Forest. Also, I have used Cross-validation to find the most optimum hyper-parameters. Finally, I compare the models and see how well each model perform using Classification metrics: Confusion Matrix, Precision, Recall and F-1 Score.

Motivation:

A fault may be defined as an unacceptable difference of at least one characteristic property or attribute of a system from acceptable usual typical performance. The main purpose of any fault diagnosis system is to determine the location and occurrence time of possible faults on the basis of accessible data and knowledge about the performance of diagnosed processes. Manual fault diagnosis system is the traditional way where an expert with electronic meter tries to obtain some information about relevant operational equipment, check the maintenance manual and then diagnose the probable causes of a particular fault. However, intelligent fault diagnosis techniques can provide quick and correct systems that help to keep product quality problems at bay and facilitates precautionary maintenance.

This project evaluates the performances of several popular and effective data mining models to diagnose seven commonly occurring faults of the steel plate namely; Pastry, Z_Scratch, K_Scatch, Stains, Dirtiness, Bumps and Other_Faults.

Data Collection:

URL : [http://archive.ics.uci.edu/ml/datasets/steel%2Bplates%2Bfaults]

Source:

Semeion, Research Center of Sciences of Communication, Via Sersale 117, 00128, Rome, Italy. www.semeion.it

References:

M Buscema, S Terzi, W Tastle, A New Meta-Classifier,in NAFIPS 2010, Toronto (CANADA),26-28 July 2010, 978-1-4244-7858-6/10 ©2010 IEEE

M Buscema, MetaNet: The Theory of Independent Judges, in Substance Use & Misuse, 33(2), 439-461,1998

Workflow :

1. Exploratory Data Analysis

   • 1.1 Count Plot to show the number of defects belonging to each defect type.
   • 1.2 Visual analysis of the correlation between the factors.
   • 1.3 Correlation Matrix

2. Model Development and Classification

   • 2.1 Data Preparation
   • 2.2 Model Development
     • 2.2.1 Logistic Regression
     • 2.2.2 KNN
     • 2.2.3 Naïve Bayes
     • 2.2.4 Decision Trees
     • 2.2.5 Random Forest

3. Principal Component Analysis

   • 3.1 PCA Transformation
     • 3.1.1 Obtain CoVariance Matrix
     • 3.1.2 Obtain EigenVectors and EigenValues
     • 3.1.3 Plot EigenValues
   • 3.2 Dimension Reduction to Principal Components
     • 3.2.1 Reduction of Dimension
   • 3.3 Modelling and Classification
     • 3.3.1 Logistic Regression
     • 3.3.2 K Nearest Neighbor
     • Naïve Bayes
     • Decision Trees
     • Random Forest

Results:

1. EDA helps in giving a preliminary glimpse on how various factors are affecting the development of Defects.
   • We notice that 'other defects', 'bumps' and 'K_Scratch' are major defects.
2. Cross-Validation is an effective way to find the best Hyper-parameters.
3. Decision Tree outperforms other Non-Ensamble models.
4. Naïve Bayes performs better on a PCA model i.e., when the dimensions are reduced.
5. Random Forests can substantially increase the efficiency but are computationally expansive when dealing with a large dataset as in this case.

Model	Classification Accuracy	Classification Error
Logistic Regression	67.61	32.39
KNN	71.46	28.54
Naïve Bayes	34.71	65.29
Decision Trees	72.04	27.96
Random Forest	79.43	20.57

PCA Model	PCA Classification Accuracy	PCA Classification Error
Logistic Regression	37.56	62.44
KNN	45.79	54.21
Naïve Bayes	43.4	56.6
Decision Trees	35.85	64.15
Random Forest	44.94	55.06

Conclusion:

Several Classification techniques fair pretty well in predicting the defect-type. Although data mining techniques are capable of extracting patterns and relationships hidden deep into large datasets, without the cooperation and feedback from the experts and professional, their results are useless. The patterns found via data mining techniques should be evaluated by professionals who have years of experience in Predicting steel plates faults.