Master thesis: Using Machine Learning Methods for Evaluating the Performance Metrics

24
An Academic presentation by Dr. Nancy Agnes, Head, Technical Operations, Tutors India Group www.tutorsindia.com Email: [email protected] PERFORMANCE EVALUATION METRICS FOR MACHINE- LEARNING BASED DISSERTATION

description

Evaluation metric plays an important role in obtaining the best possible classifier in the classification training. Thus, choosing an appropriate evaluation metric is an essential key for obtaining a selective and best possible classifier. The associated evaluation metrics have been reviewed systematically that are specifically considered as a discriminator for optimizing a classifier. In general, many possible classifiers use accuracy as a measure to classify the optimal solution during the classification evaluation. Thus, the measurement device that measures the performance of a classifier is considered as the evaluation metric. Different metrics are used to evaluate various characteristics of the classifier induced by the classification method. Contact: www.tutorsindia.com [email protected] (WA): +91-8754446690 (UK): +44-1143520021

Transcript of Master thesis: Using Machine Learning Methods for Evaluating the Performance Metrics

Page 1: Master thesis: Using Machine Learning Methods for Evaluating the Performance Metrics

An Academic presentation by Dr. Nancy Agnes, Head, Technical Operations, Tutors India Group  www.tutorsindia.comEmail: [email protected]

PERFORMANCE EVALUATIONMETRICS FOR MACHINE-LEARNING BASEDDISSERTATION

Page 2: Master thesis: Using Machine Learning Methods for Evaluating the Performance Metrics

Abstract

Introduction

Evaluation of Machine Learning

Performance measures of ML

Bayesian Inference

Recommended Algorithms

Future Topics

Conclusion

OUTLINE

Today's Discussion

Page 3: Master thesis: Using Machine Learning Methods for Evaluating the Performance Metrics

Abstract Evaluation metric plays an important role in obtaining the bestpossible classifier in the classification training.

Thus, choosing an appropriate evaluation metric is anessential key for obtaining a selective and best possibleclassifier.

The associated evaluation metrics have been reviewedsystematically that are specifically considered as a discriminatorfor optimizing a classifier.

In general, many possible classifiers use accuracy as a measureto classify the optimal solution during the classificationevaluation.

Contd...

Page 4: Master thesis: Using Machine Learning Methods for Evaluating the Performance Metrics

Thus, the measurement device that measures the performance of a classifier isconsidered as the evaluation metric.

Different metrics are used to evaluate various characteristics of the classifier inducedby the classification method.

Contd...

Page 5: Master thesis: Using Machine Learning Methods for Evaluating the Performance Metrics

Contd...

IntroductionAn important aspect of the Machine Learning process isperformance evaluation.

The right choice of performance metrics is one of the mostsignificant issues in evaluating performances.

It is also a complex task. Therefore, it should be performedcautiously in order for the machine learning application to bereliable.

Accuracy is used to assess the predictive capability of amodel on the testing samples.

Page 6: Master thesis: Using Machine Learning Methods for Evaluating the Performance Metrics

Machine learning and data mining are the fields that use this major metric.

Another alternate metric that has been used in pattern recognition and machinelearning is the ROC curve.

Thus, there are many performance metrics that have been developed for assessingthe performance of ML algorithms. 1

Page 7: Master thesis: Using Machine Learning Methods for Evaluating the Performance Metrics

EvaluationofMachineLearning

The evaluation of categorized tasks is usually done by dividingthe data set into a training data set and a testing data set.

The machine learning method is then trained on the first setof data, while the testing data set calculates the performanceindicators to assess the quality of the algorithm.

ML algorithm’s common issue lies in accessing the limitedtesting and training data.

Thus, overfitting can be a serious issue when assessing theseprograms. In order to tackle this problem, a common methodis, to employ an X-Fold Cross-Validation.

Contd...

Page 8: Master thesis: Using Machine Learning Methods for Evaluating the Performance Metrics

The cross-Validation method describes the process of dividing the entire data setinto X parts and employing each set consecutively as the test data set while mergingthe other sets to the training data.

Then the performance indicators are normalized overall validation processes.

There is no ideal performance indicator for every topic that concerns the evaluationof machine learning algorithms since every method has its own flaws andadvantages. 3

Contd...

Page 9: Master thesis: Using Machine Learning Methods for Evaluating the Performance Metrics

Image source: Evaluating Learning Algorithms 8

Page 10: Master thesis: Using Machine Learning Methods for Evaluating the Performance Metrics

Performancemeasures ofML

The performance of a classification problem can bemeasured easily using this metric.

Here, the output can be of two or more classes. Aconfusion matrix is a table with two dimensions i.e.,“Actual” and “Predicted” and also, both the dimensionshave “True Positives (TP)”, “True Negatives (TN)”, “FalsePositives (FP)”, “False Negatives (FN)”

A. CONFUSION MATRIX

Contd...

Page 11: Master thesis: Using Machine Learning Methods for Evaluating the Performance Metrics

Contd...

Page 12: Master thesis: Using Machine Learning Methods for Evaluating the Performance Metrics

Accuracy is a metric to measure the accuracy of the model.

Accuracy = Correct Predictions / Total Predictions

Accuracy is the simplest performance metric.

B. ACCURACY

Contd...

Page 13: Master thesis: Using Machine Learning Methods for Evaluating the Performance Metrics

Precision is the ratio of True Positives (TP) and the total positive predictions.

The recall is a True Positive Rate. All the positive points that are predicted positive areexplained here.

The mean of precision and recall is termed as F measure.

C. PRECISION & RECALL

Contd...

Page 14: Master thesis: Using Machine Learning Methods for Evaluating the Performance Metrics

ROC is a plot between True Positive Rate and False Positive Rate that is estimated bytaking several threshold values of probability scores from the reverse sorted list givenby a model.

D. ROC & AUC

Page 15: Master thesis: Using Machine Learning Methods for Evaluating the Performance Metrics

The recent development in machine learning has led many ITprofessionals to focus mainly on accelerating associatedworkloads, especially in machine learning.

However, in the case of unsupervised learning, the Bayesianmethod often works better than machine learning with a limited orunlabelled data set, and can influence informative priors, and alsohave interpretable approaches.

Bayesian inference model has become the most popular andaccepted model over the years as it is a huge compliment tomachine learning.

BayesianInference

Contd...

Page 16: Master thesis: Using Machine Learning Methods for Evaluating the Performance Metrics

Some recent revolutionizing research in machine learning accepts Bayesiantechniques like generative Bayesian neural networks (BNN), adversarial networks(GAN), and variational autoencoder.

Page 17: Master thesis: Using Machine Learning Methods for Evaluating the Performance Metrics

Contd...

Through visual assessment, it has been proved thatnaive Bayes was the most successful algorithm forevaluating programming performance.

Many detailed analyses were carried out statistically tofind out if there were any considerable differencesbetween the estimated accuracy of each of thealgorithms.

This is important as involved parties may prefer forchoosing an algorithm that they would like to executeand must know if the use of such algorithm(s) wouldresult in a significantly lower performance evaluation.

RecommendedAlgorithms

Page 18: Master thesis: Using Machine Learning Methods for Evaluating the Performance Metrics

The analysis identified that all of the ML algorithms, naive Bayes had comparablybest performance evaluation and thus could be used to assess the performance ofML dissertation.

Naive Bayes has been recommended as the best choice for predicting programperformance. 5

Page 19: Master thesis: Using Machine Learning Methods for Evaluating the Performance Metrics

Performance measurement has become an emergingfield during the last decades.

Organizations have many motives for usingperformance measures but the most crucial one wouldbe that they increase productivity when utilized properly.

a technique to support performance enhancement inindustrial operations.

1. EVALUATING AND MODIFYING PERFORMANCEMEASUREMENT SYSTEMS.

2. PERFORMANCE ENHANCEMENT

Future Topics

Contd...

Page 20: Master thesis: Using Machine Learning Methods for Evaluating the Performance Metrics

The main of this research is to: Build and assess a method that supportsperformance enhancement in industrial operations.

This is performed through many case studies and literature research.

The outcome is a systematically evaluated method for Performance Improvement.

prioritizing performance measures

The main aim is to decrease costs and boost the profitability of organizations tothrive in the market of competition.

3. DETERMINING PERFORMANCE MEASURES OF THE SUPPLY CHAIN

Contd...

Page 21: Master thesis: Using Machine Learning Methods for Evaluating the Performance Metrics

Many organizations use the performance measurement (PM) method to supportoperational management and strategic management processes.

This is chiefly important as it leads to modifications in organization strategy and PMsystems.

Approaches are dynamic naturally, while the current measurement systems arepredictable and stable.

Merging strategies with measurement methods is absurd and has created issues fororganizations as the strategic framework modifies.

4. A CURRENT STATE ANALYSIS TECHNIQUE FOR PERFORMANCEMEASUREMENT METHODS.

5. DYNAMIC PERFORMANCE MEASUREMENT METHODS: A FRAMEWORK FORORGANIZATIONS

Page 22: Master thesis: Using Machine Learning Methods for Evaluating the Performance Metrics

Contd...

Improving the evaluation performance of an emergingworkload, the most proficient way is to make use of existingsystems.

Another important research implemented is generic Bayesianframeworks for GPUs.

As of now, Bayesian inference is considered the bestcombination of algorithm and hardware platform forperformance evaluation.

Performance evaluation aims to approximate thegeneralization accuracy of a model in future unknown data.

Conclusion

Page 23: Master thesis: Using Machine Learning Methods for Evaluating the Performance Metrics

Contd...

In future research, research work can be carried out to improve the evaluation metricseven further.

It would be better to test those metrics on various Machine Learning cloud services to assess the services, to check how easy it is to use the metrics, and what type ofdata can be obtained using the metrics.

Research work must be carried out in this direction to build a framework that wouldhelp in prioritizing the metrics and identify a set of conditions to join results fromvarious metrics. 6

Page 24: Master thesis: Using Machine Learning Methods for Evaluating the Performance Metrics

CONTACT US

+44-1143520021UNITED KINGDOM

+91-4448137070

EMAIL

INDIA

[email protected]