Data Analytics and Machine Learning Cheng Zihan Hor Jasrene - PowerPoint PPT Presentation

Data Analytics and Machine Learning Cheng Zihan Hor Jasrene Joshua Tan EEE03

Content Outline 1. 1. Int ntrodu ductio ction 2. 2. Aims & Objec ectiv tives es 3. Method 3. odolo logy gy 4. 4. Results ults & Discus cussion sion 5. 5. Conclusion usion 2

1. Introduction What is Machine Learning?

What is Machine Learning? ▰ Application of AI ▰ Identifying patterns & make decisions based on past data ▰ Analyse & interpret trends to generate new insights ▰ Handle massive data without human intervention 4

Applications ▰ Online Fraud Detection ▰ Malware Filtering ▰ Facial Recognition ▰ Virtual Personal Assistants 5

Classifications of Machine Learning Algorithms Supervised Learning Unsupervised Learning ▰ Past st data ta to train ain alg lgori rithm hm ▰ Information to train data is not classified ▰ Compa mpare re predicted dicted wit ith h exp xpect ected d outpu tput ▰ Model opti timis isation tion to reduce ce error ▰ Cla lassif sificat ication ion & Regr gression ession 6

2. Aims & Objectives Why did we work on this project?

Aims & Objectives ▰ For Perceptron Learning Algorithm & Adaline Gradient Descent Algorithm to undergo supervised learning ▰ Compare effectiveness of algorithms to classify data 8

3. Methodology How did we carry out the project?

Data Sets ▰ Iris.data set for analysis by algorithms ○ 3 classes of 50 instances each ▰ 5 Attributes ○ Sepal Length ○ Sepal Width ○ Petal Length ○ Petal Width ○ Type of Iris Plant 10

Visual representation of data training set 11

Perceptron Updates using individual instances of data 12

Perceptron Learning Rule ▰ Perceptron ron trained d with past data classifi sified ed into matrice ices ▰ Aggregates tes input based d on the we weights to gen ener erate e an o outpu put ▰ AIM: : Generate te weights of a perceptron on for r each attribute e accurat rately ely 13

Perceptron Learning Rule p (a) z i = w 0 + ∑ w j x i,j ; j = 1 with z being the intermediate , w the weight and x the input. (b) Limit added to value of z i to fit the output y’ i : y’ i = ⎰ 1 if z i ≥ 0; ⎱ -1 otherwise. 14

Perceptron Learning Rule (c) Update Weights of Perceptron: Calculate error e i -- difference between the output and the target class label -- and multiply it by the update step-size. e i = y i - y’ i ; w 0 := w 0 + η e i w := w + η x i e i. 15

Adaline Updates using gradient of training data 16

Similarities ▰ Classifiers ○ Iris data converted to numerical format ○ Recognisable by algorithms ▰ Weights ○ Prediction of data points ○ Continuously updated to minimise error ▰ Decision boundary ○ Classification of data by algorithms 17

Experimental Procedure Weight Algorithm Comparing Comparing generation class error decision and created graphs boundaries updates 18

4. Results & Discussion What did we find out from the project? 19

classified breeds: setosa Classification graph and vesicolor ▰ used in comparison with the classifications of the respective algorithms to determine their accuracy and sensitivities. 20

Comparison of magnitude blue: Perceptron Algorithm orange: Adaline of errors produced Algorithm lower rate of learning higher rate of learning 21

Classification by respective Algorithms Perceptron Algorithm Adaline Algorithm ▰ Perceptron Algorithm is more sensitive to outliers than Adaline Algorithm 22

Conclusion ▰ Perceptron Learning Algorithm is more sensitive and accurate in the classification of data as compared to the Adaline Algorithm. ▰ important indications on the unlimited potential in training and classifying data 23

ACKNOWLEDGEMENTS ▰ Supervisor: Prof. Andy W.H. Khong ▰ School teacher mentors: Mr Nicholas Wong Mr Lim Cher Chuan Dr Goh Ker Liang 24

THE END Thank you so much for your time! 25