Machine Learning End to End at Surat

This is presented in GDG Surat Devfest. It starts with basic introduction of Machine Learning and ends with the deployment of model using flask.

Transcript

1. Krunal Kapadiya, Volansys @krunal3kapadiya Surat Machine Learning End to End
5. Applications of Machine Learning
6. Two ways Google can help us! Surat
7. What is Machine Learning? Input Output Magic
8. What is Machine Learning? "A computer program is said to learn from experience E with respect to some class of tasks T and performance measure P if its performance at tasks in T, as measured by P, improves with experience E.” - Tom. M. Mitchell
9. Proprietary + Confidential
11. Set Goal Split Data Train Model Results Run an ML Experiment Steps to follow Test Model Surat
12. Training and splitting data with validations Steps to follow 20% Test case 80% Training set Total numbers of training set
13. Training and splitting data with validations Steps to follow 15% Test set 70% Training set Total numbers of training set 15% Validation
14. Find apple or orange problem Traditional approach to solve problem
15. Find apple or orange problem Traditional approach to solve problem
16. Find apple or orange problem Traditional approach to solve problem
17. Weight Texture Label 150g Bumpy Orange 170g Bumpy Orange 140g Smooth Apple 130g Smooth Apple Feature Feature Find apple or orange problem Training Data
18. Weight Texture Label 150g Bumpy Orange 170g Bumpy Orange 140g Smooth Apple 130g Smooth Apple Feature Feature Examples Find apple or orange problem Training Data
19. Orange Apple Weight = 150 G Yes No Yes No Texture = bumpy ? ... Decision Tree Find apple or orange problem
21. Types of Machine Learning Surat
22. Types of Machine learning Based on Human supervision - Supervised - Unsupervised - Reinforcement
23. Other types of Machine learning Based on learning Online learning Offline learning Based on Data patterns Instance based Model based
24. Supervised model
25. Algorithms in supervised learning - K Nearest Neighbours - Linear regressions - Logistic regressions - Support vector machines - Decision tree and random forests
26. Show me code Surat
27. # Select a linear model model = sklearn.linear_model.LinearRegression() # Train the model model.fit(X, y) # Make a prediction for Cyprus X_new = [[22587]] # Cyprus' GDP per capita print(model.predict(X_new)) # outputs [[ 5.96242338]]
28. Unsupervised model
29. Algorithms in unsupervised learning - Clustering - k-Means - Hierarchical cluster analysis (HCA) - Association rule learning - Apriori - Eclat
30. Problems: Customer segmentation We have data, it is in csv format having rows - Customer Id - Gender - Age - Annual Income - Spending score
31. kmeans = KMeans(n_clusters=5, init='k-means++', random_state=0) y_kmeans = kmeans.fit_predict(X) # will return the predicted class
32. Problems: Customer segmentation
33. Reinforcement learning
34. Algorithms in Reinforcement learning - Q-learning - SARSA - DQN - DDPG - Multi arm bandit problem
35. Neural network
36. Convolutional Neural Network
37. - Pooling
38. Show me code Surat
39. x = tf.reshape(x, shape=[-1, 28, 28, 1]) # Convolution Layer with 32 filters and a kernel size of 5 conv1 = tf.layers.conv2d(x, 32, 5, activation=tf.nn.relu) # Max Pooling (down-sampling) with strides of 2 and kernel size of 2 conv1 = tf.layers.max_pooling2d(conv1, 2, 2) # Convolution Layer with 64 filters and a kernel size of 3 conv2 = tf.layers.conv2d(conv1, 64, 3, activation=tf.nn.relu) # Max Pooling (down-sampling) with strides of 2 and kernel size of 2 conv2 = tf.layers.max_pooling2d(conv2, 2, 2)
40. # Flatten the data to a 1-D vector for the fully connected layer fc1 = tf.contrib.layers.flatten(conv2) # Fully connected layer (in tf contrib folder for now) fc1 = tf.layers.dense(fc1, 1024) # Apply Dropout (if is_training is False, dropout is not applied) fc1 = tf.layers.dropout(fc1, rate=dropout, training=is_training) # Output layer, class prediction out = tf.layers.dense(fc1, n_classes)
41. input_fn = tf.estimator.inputs.numpy_input_fn( x={'images': mnist.train.images}, y=mnist.train.labels, batch_size=batch_size, num_epochs=None, shuffle=True) # Train the Model model.train(input_fn, steps=num_steps)
42. But, real problems are... - Insufficient quantity of training data - Non representative training data - Poor quality data - Irrelevant features - Overfitting - Underfitting
43. Lastly, how to publish model in production Surat
44. Add ML model in production using
45. header = {'Content-Type': 'application/json', \ 'Accept': 'application/json'} """Reading test batch """ df = pd.read_csv('../data/test.csv', encoding="utf-8-sig") df = df.head() """Converting Pandas Dataframe to json """ data = df.to_json(orient='records') Add ML model in production using flask Source: https://www.analyticsvidhya.com/blog/2017/09/machine-learning-models-as-apis-using-flask/
46. //OUTPUT [ { "Loan_ID": "LP001015", "Gender": "Male", "Married": "Yes", "Dependents": "0", "Education": "Graduate", "Self_Employed": "No", "ApplicantIncome": 5720, "CoapplicantIncome": 0, "LoanAmount": 110, "Loan_Amount_Term": 360, "Credit_History": 1, "Property_Area": "Urban" } ] Add ML model in production using flask Source: https://www.analyticsvidhya.com/blog/2017/09/machine-learning-models-as-apis-using-flask/
47. How can I start it - Look at the dataset - Write down columns and it’s correlation - Make questions derived from the dataset - Explanatory Analysis with visualization - Frame problem - Create solution by creating model
48. Where to start learning ML • Data analytics vidhya • KDnuggets • Coursera and Udacity course
49. Start learning by own • https://developers.google.com/machine-learning/crash-course/ml-intro • https://www.kaggle.com/learn/overview • https://www.tensorflow.org/tutorials/ • https://pandas.pydata.org/pandas-docs/stable/10min.html
50. ML Wordtoit is available in Assistant Link:- http://bit.ly/mlwordtoit
51. Krunal Kapadiya, Volansys @krunal3kapadiya Surat Thank you! Q&A #IndiaMLCC