title
Introduction to Decision Trees

description

detail
{'title': 'Introduction to Decision Trees', 'heatmap': [{'end': 168.639, 'start': 135.546, 'weight': 0.803}, {'end': 636.678, 'start': 614.952, 'weight': 0.857}, {'end': 780.524, 'start': 762.409, 'weight': 0.796}, {'end': 1267.977, 'start': 1245.396, 'weight': 0.83}, {'end': 1363.693, 'start': 1282.239, 'weight': 0.777}, {'end': 1459.36, 'start': 1436.799, 'weight': 0.746}], 'summary': 'Introduces decision trees as a non-linear function, explaining their structure as a tree classifier with decision and leaf nodes, their usage for classification and regression, application in loan approval and car mileage prediction, learning process from training examples, and their versatility in representing boolean functions.', 'chapters': [{'end': 229.381, 'segs': [{'end': 201.322, 'src': 'heatmap', 'start': 135.546, 'weight': 0, 'content': [{'end': 147.332, 'text': 'So, in decision nodes they specify a choice or a test based on this you can decide which direction you can go.', 'start': 135.546, 'duration': 11.786}, {'end': 157.897, 'text': 'So, in a decision tree, we test something and that test may have more than one result, and based on the value of this test,', 'start': 147.712, 'duration': 10.185}, {'end': 160.458, 'text': 'you either follow this branch or this branch.', 'start': 157.897, 'duration': 2.561}, {'end': 168.639, 'text': 'So, this test is usually done on the value of a feature or attribute of the instance.', 'start': 161.693, 'duration': 6.946}, {'end': 174.705, 'text': 'So, test is on some attribute and there is a branch for each outcome.', 'start': 170.361, 'duration': 4.344}, {'end': 180.19, 'text': 'So, there may be two outcomes or in some cases you can have more than two outcomes.', 'start': 175.065, 'duration': 5.125}, {'end': 182.612, 'text': 'And then there are leaf nodes.', 'start': 181.091, 'duration': 1.521}, {'end': 187.497, 'text': 'So, leaf node indicates the classification of an example.', 'start': 183.894, 'duration': 3.603}, {'end': 194.936, 'text': 'or the value of the example.', 'start': 193.354, 'duration': 1.582}, {'end': 201.322, 'text': 'Decision trees can be used both for classification and regression.', 'start': 196.137, 'duration': 5.185}], 'summary': 'Decision trees use tests on features to classify examples or predict values, with multiple possible outcomes and leaf nodes for classification or regression.', 'duration': 112.714, 'max_score': 135.546, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/FuJVLsZYkuE/pics/FuJVLsZYkuE135546.jpg'}], 'start': 19.778, 'title': 'Decision trees overview', 'summary': 'Introduces decision trees as a non-linear function in the module introduction to linear regression and decision trees, explaining its structure as a tree classifier with decision and leaf nodes. it also explains the structure and usage of decision trees, including their ability to test for multiple outcomes, their use for both classification and regression, and the process of navigating from the root to leaf nodes to determine the value of an example.', 'chapters': [{'end': 128.652, 'start': 19.778, 'title': 'Introduction to decision trees', 'summary': 'Introduces decision trees as a non-linear function in the module introduction to linear regression and decision trees, explaining its structure as a tree classifier with decision and leaf nodes.', 'duration': 108.874, 'highlights': ["Decision trees are a non-linear function and a tree-structured classifier with decision and leaf nodes. The learning algorithm introduced in today's lecture is a decision tree, which is a non-linear function and a tree-structured classifier with decision and leaf nodes.", 'A decision tree is a tree classifier with nodes and branches, including decision nodes and leaf nodes. A decision tree is a tree-structured classifier with nodes and branches, consisting of decision nodes and leaf nodes.', 'Definition of a decision tree as a classifier with tree structure and two types of nodes. A decision tree is defined as a classifier with a tree structure and two types of nodes: decision nodes and leaf nodes.']}, {'end': 229.381, 'start': 135.546, 'title': 'Decision trees overview', 'summary': 'Explains the structure and usage of decision trees, including their ability to test for multiple outcomes, their use for both classification and regression, and the process of navigating from the root to leaf nodes to determine the value of an example.', 'duration': 93.835, 'highlights': ['Decision trees can be used for both classification and regression, with a primary focus on classification, and they test for multiple outcomes based on the value of a feature or attribute, which determines the corresponding branch to follow.', 'The process of navigating a decision tree involves starting at the root and moving through the branches based on test values until reaching a leaf node, where the value of the example is determined.', 'Leaf nodes in decision trees indicate the classification of an example or the value of the example, providing clarity in the final outcome.']}], 'duration': 209.603, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/FuJVLsZYkuE/pics/FuJVLsZYkuE19778.jpg', 'highlights': ['Decision trees are a non-linear function and a tree-structured classifier with decision and leaf nodes.', 'Decision trees can be used for both classification and regression, with a primary focus on classification.', 'The process of navigating a decision tree involves starting at the root and moving through the branches based on test values until reaching a leaf node.']}, {'end': 569.918, 'segs': [{'end': 366.639, 'src': 'embed', 'start': 319.91, 'weight': 0, 'content': [{'end': 339.61, 'text': 'If the applicant is employed, then you have another test you check the income of the applicant and if the income is high, you approve the loan,', 'start': 319.91, 'duration': 19.7}, {'end': 342.891, 'text': 'and if the income is low, you reject the loan.', 'start': 339.61, 'duration': 3.281}, {'end': 349.633, 'text': 'So, this is an example of a decision tree.', 'start': 346.672, 'duration': 2.961}, {'end': 355.095, 'text': 'We have 3 decision loads and 4 relief loads.', 'start': 350.574, 'duration': 4.521}, {'end': 366.639, 'text': 'Now, how do you use the decision tree? Suppose that applicant is employed, has low income.', 'start': 357.202, 'duration': 9.437}], 'summary': "Decision tree: 3 decision loads, 4 relief loads, based on applicant's income for loan approval.", 'duration': 46.729, 'max_score': 319.91, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/FuJVLsZYkuE/pics/FuJVLsZYkuE319910.jpg'}, {'end': 427.556, 'src': 'embed', 'start': 400.101, 'weight': 1, 'content': [{'end': 408.924, 'text': 'you have the different attributes of the applicant, including the income of the applicant, whether the applicant is employed,', 'start': 400.101, 'duration': 8.823}, {'end': 413.705, 'text': 'what is the credit score and several other attributes of the applicants are there,', 'start': 408.924, 'duration': 4.781}, {'end': 421.635, 'text': 'and also the suggested action whether the loan should be approved or rejected.', 'start': 413.705, 'duration': 7.93}, {'end': 423.235, 'text': 'that is given in the training set.', 'start': 421.635, 'duration': 1.6}, {'end': 427.556, 'text': 'From the training set you can come up with a decision tree like this.', 'start': 423.715, 'duration': 3.841}], 'summary': 'Using applicant attributes, a decision tree can be created to suggest loan approval or rejection.', 'duration': 27.455, 'max_score': 400.101, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/FuJVLsZYkuE/pics/FuJVLsZYkuE400101.jpg'}, {'end': 531.902, 'src': 'embed', 'start': 494.607, 'weight': 2, 'content': [{'end': 503.211, 'text': 'So, this is how you can also accommodate continuous variables as attributes in a decision tree.', 'start': 494.607, 'duration': 8.604}, {'end': 512.456, 'text': 'Now, suppose if the age is less than 30, you have another decision variable checking if the applicant is a student.', 'start': 504.192, 'duration': 8.264}, {'end': 524.318, 'text': 'If he is a student then let us say we say that he is likely to buy a computer.', 'start': 514.397, 'duration': 9.921}, {'end': 531.902, 'text': 'So, this is the decision tree whether an applicant whether a person is likely to buy a computer.', 'start': 524.898, 'duration': 7.004}], 'summary': 'A decision tree model predicts computer purchase based on age and student status.', 'duration': 37.295, 'max_score': 494.607, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/FuJVLsZYkuE/pics/FuJVLsZYkuE494607.jpg'}], 'start': 229.541, 'title': 'Decision trees in loan approval', 'summary': 'Discusses the application of decision trees in loan approval, demonstrating with examples of decision and leaf nodes, and using factors like employment status, income, credit score, and age to determine approval or rejection, and likelihood of computer purchase.', 'chapters': [{'end': 400.101, 'start': 229.541, 'title': 'Decision trees for loan approval', 'summary': 'Discusses decision trees and their application in loan approval, illustrating the decision tree with an example of 3 decision nodes and 4 leaf nodes, and demonstrates how to use the decision tree to determine loan approval based on employment status and income.', 'duration': 170.56, 'highlights': ['The decision tree example consists of 3 decision nodes and 4 leaf nodes, demonstrating the process of determining loan approval based on employment status, credit score, and income.', 'The decision tree is used to determine loan approval based on specific criteria, such as employment status and income, leading to a clear decision on whether to approve or reject the loan for a given applicant.', "The chapter provides a practical example of using a decision tree to assess loan approval, showcasing how the decision tree is utilized to make a definitive decision based on the applicant's employment and income status."]}, {'end': 461.01, 'start': 400.101, 'title': 'Decision tree for applicant attributes', 'summary': "Discusses using decision trees to analyze applicant attributes, including income, employment status, credit score, and other factors, to determine loan approval or rejection, and also mentions using decision trees to predict a person's likelihood to buy a computer based on their age as a continuous valued attribute.", 'duration': 60.909, 'highlights': ['Using decision trees to analyze applicant attributes for loan approval or rejection The training set includes attributes such as income, employment status, credit score, and others to determine loan approval or rejection.', "Using decision trees to predict a person's likelihood to buy a computer based on age Mentions using decision trees to predict a person's likelihood to buy a computer based on their age as a continuous valued attribute."]}, {'end': 569.918, 'start': 461.83, 'title': 'Decision tree for predicting computer purchase', 'summary': 'Discusses using decision trees to predict computer purchase based on age, student status, and credit rating, with examples of age classes and corresponding purchasing likelihood.', 'duration': 108.088, 'highlights': ['Using decision trees to predict computer purchase based on age, student status, and credit rating The chapter explains the use of decision trees to predict computer purchase based on various attributes such as age, student status, and credit rating.', 'Example of age classes and corresponding purchasing likelihood The example illustrates dividing age into classes and determining purchasing likelihood based on age ranges, such as age less than 30, between 30 to 40, and greater than 40.', 'Consideration of student status and its impact on purchasing likelihood The discussion involves considering the status of being a student as a decision variable affecting the likelihood of purchasing a computer, where students are likely to buy a computer.', 'Incorporating credit rating into the decision tree The transcript shows the incorporation of credit rating as a factor in the decision tree, indicating that individuals with fair credit ratings are likely to buy a computer, while those with excellent credit ratings are not likely to buy.']}], 'duration': 340.377, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/FuJVLsZYkuE/pics/FuJVLsZYkuE229541.jpg', 'highlights': ['The decision tree example consists of 3 decision nodes and 4 leaf nodes, demonstrating the process of determining loan approval based on employment status, credit score, and income.', 'Using decision trees to analyze applicant attributes for loan approval or rejection The training set includes attributes such as income, employment status, credit score, and others to determine loan approval or rejection.', 'Using decision trees to predict computer purchase based on age, student status, and credit rating The chapter explains the use of decision trees to predict computer purchase based on various attributes such as age, student status, and credit rating.']}, {'end': 818.494, 'segs': [{'end': 602.058, 'src': 'embed', 'start': 570.338, 'weight': 0, 'content': [{'end': 572.939, 'text': 'This is another example of a decision tree.', 'start': 570.338, 'duration': 2.601}, {'end': 585.043, 'text': 'And this is another example in this slide here, this is a decision tree to predict the car mileage prediction.', 'start': 575.56, 'duration': 9.483}, {'end': 588.233, 'text': 'Is the weight of the car heavy? Yes.', 'start': 586.252, 'duration': 1.981}, {'end': 589.473, 'text': 'Then high mileage.', 'start': 588.513, 'duration': 0.96}, {'end': 593.014, 'text': 'Is it no? Then check the horse power.', 'start': 590.113, 'duration': 2.901}, {'end': 597.496, 'text': 'If horse power is less than equal to 86, then high mileage.', 'start': 593.455, 'duration': 4.041}, {'end': 599.037, 'text': 'If no, low mileage.', 'start': 597.896, 'duration': 1.141}, {'end': 602.058, 'text': 'This is another example of a decision tree.', 'start': 599.497, 'duration': 2.561}], 'summary': 'Example of decision tree for car mileage prediction.', 'duration': 31.72, 'max_score': 570.338, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/FuJVLsZYkuE/pics/FuJVLsZYkuE570338.jpg'}, {'end': 636.678, 'src': 'heatmap', 'start': 614.952, 'weight': 0.857, 'content': [{'end': 627.627, 'text': 'Now, given some training examples, it is possible that there can be many decision trees which fit the training example.', 'start': 614.952, 'duration': 12.675}, {'end': 636.678, 'text': 'then our question is which decision tree should we choose among the many possible decision trees.', 'start': 629.628, 'duration': 7.05}], 'summary': 'Multiple decision trees can fit training examples, prompting the choice of an optimal tree.', 'duration': 21.726, 'max_score': 614.952, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/FuJVLsZYkuE/pics/FuJVLsZYkuE614952.jpg'}, {'end': 818.494, 'src': 'heatmap', 'start': 762.409, 'weight': 1, 'content': [{'end': 766.412, 'text': 'Therefore, we look for some greedy algorithms,', 'start': 762.409, 'duration': 4.003}, {'end': 780.524, 'text': 'we search for a good tree and we have to decide how we can come up with a good tree for learning the decision tree.', 'start': 766.412, 'duration': 14.112}, {'end': 788.01, 'text': 'Now, this is some example data on which we can use a decision tree and so these are some training examples.', 'start': 781.244, 'duration': 6.766}, {'end': 796.033, 'text': 'these are certain attributes author, thread, length, where.', 'start': 789.764, 'duration': 6.269}, {'end': 802.843, 'text': 'So, you want to know whether a user reads a thread or skips a thread.', 'start': 796.414, 'duration': 6.429}, {'end': 810.065, 'text': 'And given the attributes, who is the author of the post, whether the thread is new or old,', 'start': 804.159, 'duration': 5.906}, {'end': 814.189, 'text': 'what is the length of the post and where the user currently is?', 'start': 810.065, 'duration': 4.124}, {'end': 818.494, 'text': 'you want to decide the action of the user skips or reads.', 'start': 814.189, 'duration': 4.305}], 'summary': 'Using greedy algorithms to learn a decision tree for predicting user actions based on attributes like author, thread, and post length.', 'duration': 131.178, 'max_score': 762.409, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/FuJVLsZYkuE/pics/FuJVLsZYkuE762409.jpg'}], 'start': 570.338, 'title': 'Decision trees in car mileage prediction', 'summary': 'Covers the use of decision trees to predict car mileage based on weight and horsepower, and discusses bias and preference for smaller trees, highlighting challenges in finding the smallest tree.', 'chapters': [{'end': 660.116, 'start': 570.338, 'title': 'Decision tree for car mileage prediction', 'summary': 'Discusses the use of decision trees to predict car mileage, highlighting the criteria for high and low mileage based on car weight and horsepower, and the process of choosing the best decision tree among multiple possibilities.', 'duration': 89.778, 'highlights': ['The process of using a decision tree to predict car mileage based on weight and horsepower is explained, determining high or low mileage based on specific criteria.', 'The challenge of selecting the best decision tree among multiple possibilities is discussed, considering the potential presence of noisy data and the need to minimize errors.']}, {'end': 818.494, 'start': 660.116, 'title': 'Decision tree bias and preference', 'summary': 'Discusses the bias and preference in decision trees, emphasizing the preference for smaller trees with fewer nodes or smaller depth, and the challenges in finding the smallest decision tree that fits the data.', 'duration': 158.378, 'highlights': ['The preference for smaller decision trees is commonly expressed as a bias, aiming for trees with smaller depth or fewer nodes, which restricts the hypothesis space (Week 1).', 'Finding the smallest decision tree that fits the data is a computationally hard problem, leading to the exploration of greedy algorithms to search for a good tree in the space of decision trees.', 'Attributes such as author, thread, length, and user location are used to determine whether a user reads or skips a thread, illustrating the practical application of decision trees for action prediction.']}], 'duration': 248.156, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/FuJVLsZYkuE/pics/FuJVLsZYkuE570338.jpg', 'highlights': ['The process of using a decision tree to predict car mileage based on weight and horsepower is explained, determining high or low mileage based on specific criteria.', 'The challenge of selecting the best decision tree among multiple possibilities is discussed, considering the potential presence of noisy data and the need to minimize errors.', 'Attributes such as author, thread, length, and user location are used to determine whether a user reads or skips a thread, illustrating the practical application of decision trees for action prediction.', 'The preference for smaller decision trees is commonly expressed as a bias, aiming for trees with smaller depth or fewer nodes, which restricts the hypothesis space (Week 1).', 'Finding the smallest decision tree that fits the data is a computationally hard problem, leading to the exploration of greedy algorithms to search for a good tree in the space of decision trees.']}, {'end': 1185.867, 'segs': [{'end': 864.398, 'src': 'embed', 'start': 819.615, 'weight': 0, 'content': [{'end': 822.578, 'text': 'So, given these attributes you want to learn a decision tree.', 'start': 819.615, 'duration': 2.963}, {'end': 834.985, 'text': 'When you get some new examples, you can find out in the case of E7 whether the reader will read or skip.', 'start': 825.141, 'duration': 9.844}, {'end': 841.828, 'text': 'Now, so let us see how we can learn decision trees.', 'start': 837.586, 'duration': 4.242}, {'end': 844.809, 'text': 'So, we are given training examples.', 'start': 843.008, 'duration': 1.801}, {'end': 847.53, 'text': 'For example, we are given this sort of training example.', 'start': 844.889, 'duration': 2.641}, {'end': 864.398, 'text': 'Suppose, D is the set of training examples.', 'start': 862.037, 'duration': 2.361}], 'summary': 'Learning decision trees from training examples to predict reader behavior.', 'duration': 44.783, 'max_score': 819.615, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/FuJVLsZYkuE/pics/FuJVLsZYkuE819615.jpg'}, {'end': 1029.646, 'src': 'embed', 'start': 998.294, 'weight': 1, 'content': [{'end': 1007.522, 'text': 'So, at every step we have to make a decision whether to stop growing the tree at that node or whether to continue.', 'start': 998.294, 'duration': 9.228}, {'end': 1014.548, 'text': 'If we want to continue growing the tree, we have to decide which attribute to split on.', 'start': 1008.103, 'duration': 6.445}, {'end': 1017.411, 'text': 'So, these are the decisions that we have to make.', 'start': 1014.969, 'duration': 2.442}, {'end': 1029.646, 'text': 'For example, on these examples, suppose we take length as the attribute and let us say the examples that are there in that node.', 'start': 1019.164, 'duration': 10.482}], 'summary': 'Decide whether to stop or continue growing the tree and which attribute to split on.', 'duration': 31.352, 'max_score': 998.294, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/FuJVLsZYkuE/pics/FuJVLsZYkuE998294.jpg'}], 'start': 819.615, 'title': 'Learning decision trees', 'summary': 'Explains the process of learning decision trees from training examples, including the attributes and recursive splitting, with examples of attribute selection and tree growth.', 'chapters': [{'end': 1185.867, 'start': 819.615, 'title': 'Learning decision trees', 'summary': 'Explains the process of learning decision trees from training examples, including the attributes and recursive splitting, with examples of attribute selection and tree growth.', 'duration': 366.252, 'highlights': ['The training examples are used to recursively build decision trees by choosing attributes for node splitting, with the aim of reaching a leaf quickly for a smaller tree.', 'Attributes such as length and thread are used for node splitting, with specific examples of the number of examples for each attribute value and the decision-making process.', 'Examples of decision trees are provided, demonstrating different approaches to reaching a leaf node, either with specific values or with dominant values.', 'The decision tree building process involves making decisions at each node, including whether to stop growing the tree and which attribute to split on, based on the examples and their attributes.']}], 'duration': 366.252, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/FuJVLsZYkuE/pics/FuJVLsZYkuE819615.jpg', 'highlights': ['The training examples are used to recursively build decision trees by choosing attributes for node splitting, with the aim of reaching a leaf quickly for a smaller tree.', 'The decision tree building process involves making decisions at each node, including whether to stop growing the tree and which attribute to split on, based on the examples and their attributes.', 'Attributes such as length and thread are used for node splitting, with specific examples of the number of examples for each attribute value and the decision-making process.', 'Examples of decision trees are provided, demonstrating different approaches to reaching a leaf node, either with specific values or with dominant values.']}, {'end': 1460.321, 'segs': [{'end': 1222.27, 'src': 'embed', 'start': 1193.146, 'weight': 1, 'content': [{'end': 1199.612, 'text': 'So, let us take one example, this example is taken by from the book on machine learning by Tom Mitchell.', 'start': 1193.146, 'duration': 6.466}, {'end': 1206.079, 'text': 'So, where he looks at a decision tree to decide whether it is a good day to play tennis.', 'start': 1200.113, 'duration': 5.966}, {'end': 1213.065, 'text': 'The attributes used in the decision tree are outlook, outlook can be sunny, overcast or rainy.', 'start': 1206.519, 'duration': 6.546}, {'end': 1222.27, 'text': 'Humidity, which has values high and normal, wind has values strong and weak, and temperature has hot, mild and cool.', 'start': 1213.906, 'duration': 8.364}], 'summary': "Example from tom mitchell's book on machine learning uses decision tree with attributes outlook, humidity, wind, and temperature.", 'duration': 29.124, 'max_score': 1193.146, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/FuJVLsZYkuE/pics/FuJVLsZYkuE1193146.jpg'}, {'end': 1280.203, 'src': 'heatmap', 'start': 1245.396, 'weight': 0, 'content': [{'end': 1259.018, 'text': 'Now in this decision tree we have internal nodes at decision nodes which test an attribute and branch corresponds to an attribute value node and there are leaves which assign a classification.', 'start': 1245.396, 'duration': 13.622}, {'end': 1267.977, 'text': 'Now, given this decision tree and given a new example for which outlook is sunny, temperature is hot, humidity is high,', 'start': 1259.833, 'duration': 8.144}, {'end': 1272.179, 'text': 'wind is weak you want to know whether it is a good day to play tennis.', 'start': 1267.977, 'duration': 4.202}, {'end': 1280.203, 'text': 'So, you first check the route it says outlook, because outlook is sunny you go to the left branch.', 'start': 1273.139, 'duration': 7.064}], 'summary': 'Decision tree uses attributes to classify examples. can determine tennis playing conditions.', 'duration': 20.37, 'max_score': 1245.396, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/FuJVLsZYkuE/pics/FuJVLsZYkuE1245396.jpg'}, {'end': 1333.592, 'src': 'embed', 'start': 1307.329, 'weight': 2, 'content': [{'end': 1316.278, 'text': 'So, decision tree is a very flexible function which can represent disjunction of conjunction and thus it can represent all Boolean function.', 'start': 1307.329, 'duration': 8.949}, {'end': 1321.463, 'text': 'If a decision tree is of sufficiently large size it can express all Boolean functions.', 'start': 1316.298, 'duration': 5.165}, {'end': 1326.767, 'text': 'Now, as we said, that the learning problem is, given a decision tree,', 'start': 1322.464, 'duration': 4.303}, {'end': 1333.592, 'text': 'we have to find a good tree and there are two choices that you have to make if you decide, at a particular point,', 'start': 1326.767, 'duration': 6.825}], 'summary': 'Decision tree is flexible, can represent all boolean functions, and learning involves finding a good tree.', 'duration': 26.263, 'max_score': 1307.329, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/FuJVLsZYkuE/pics/FuJVLsZYkuE1307329.jpg'}, {'end': 1410.021, 'src': 'heatmap', 'start': 1282.239, 'weight': 3, 'content': [{'end': 1292.145, 'text': 'then you check for humidity if humidity is high you take the left branch and then it says no it is not a good day to play tennis.', 'start': 1282.239, 'duration': 9.906}, {'end': 1296.387, 'text': 'So, you output no this is how you use a decision tree.', 'start': 1292.565, 'duration': 3.822}, {'end': 1306.768, 'text': 'A decision tree can be expressed as a Boolean function which is a disjunction of conjunctions.', 'start': 1298.76, 'duration': 8.008}, {'end': 1316.278, 'text': 'So, decision tree is a very flexible function which can represent disjunction of conjunction and thus it can represent all Boolean function.', 'start': 1307.329, 'duration': 8.949}, {'end': 1321.463, 'text': 'If a decision tree is of sufficiently large size it can express all Boolean functions.', 'start': 1316.298, 'duration': 5.165}, {'end': 1326.767, 'text': 'Now, as we said, that the learning problem is, given a decision tree,', 'start': 1322.464, 'duration': 4.303}, {'end': 1333.592, 'text': 'we have to find a good tree and there are two choices that you have to make if you decide, at a particular point,', 'start': 1326.767, 'duration': 6.825}, {'end': 1336.434, 'text': 'whether you should stop or whether you should continue.', 'start': 1333.592, 'duration': 2.842}, {'end': 1344.339, 'text': 'If you want to continue you have to choose a test that is you have to choose an attribute or a feature to continue with.', 'start': 1336.774, 'duration': 7.565}, {'end': 1350.644, 'text': 'Now, we will just give the framework of a basic decision tree algorithm.', 'start': 1344.919, 'duration': 5.725}, {'end': 1360.811, 'text': 'This algorithm is called top down induction of decision tree and this is the basic ID3 algorithm which was proposed by Quinlan.', 'start': 1351.184, 'duration': 9.627}, {'end': 1363.693, 'text': 'So, these are the steps of the algorithm.', 'start': 1361.912, 'duration': 1.781}, {'end': 1368.597, 'text': 'At the current node, you choose the best decision attribute.', 'start': 1364.574, 'duration': 4.023}, {'end': 1373.639, 'text': 'Then assign a as the decision attribute for the node.', 'start': 1370.194, 'duration': 3.445}, {'end': 1379.868, 'text': 'For each value of a that is the outcome you create a new descendant.', 'start': 1374.781, 'duration': 5.087}, {'end': 1387.481, 'text': 'and then the training examples will get split into the different branches.', 'start': 1381.576, 'duration': 5.905}, {'end': 1397.41, 'text': 'So, you sort or split the training example to the leaves of the current node according to the attribute value of the branch.', 'start': 1388.382, 'duration': 9.028}, {'end': 1405.457, 'text': 'So, at a particular node if you find that all the training examples have the same class then you can stop.', 'start': 1398.311, 'duration': 7.146}, {'end': 1410.021, 'text': 'otherwise you can again continue this process.', 'start': 1406.358, 'duration': 3.663}], 'summary': 'Decision trees can represent all boolean functions and are used to make decisions based on given attributes and features.', 'duration': 58.837, 'max_score': 1282.239, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/FuJVLsZYkuE/pics/FuJVLsZYkuE1282239.jpg'}, {'end': 1460.321, 'src': 'heatmap', 'start': 1436.799, 'weight': 5, 'content': [{'end': 1443.265, 'text': 'Once we choose a node, we have to choose the best attribute of that node and we have to decide when we want to stop.', 'start': 1436.799, 'duration': 6.466}, {'end': 1446.948, 'text': 'These are the decisions that we have to take in a decision tree.', 'start': 1443.725, 'duration': 3.223}, {'end': 1449.691, 'text': 'And in the next class.', 'start': 1447.489, 'duration': 2.202}, {'end': 1459.36, 'text': 'so these are the two decisions that we have to take, and in the next class we will look at some specific heuristics to may take a decision on this.', 'start': 1449.691, 'duration': 9.669}, {'end': 1460.321, 'text': 'Thank you.', 'start': 1459.941, 'duration': 0.38}], 'summary': 'Decision tree involves choosing best attribute and stopping criteria. specific heuristics will be discussed next class.', 'duration': 23.522, 'max_score': 1436.799, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/FuJVLsZYkuE/pics/FuJVLsZYkuE1436799.jpg'}], 'start': 1193.146, 'title': 'Decision trees for tennis', 'summary': "Discusses the concept of decision trees for determining whether it's a good day to play tennis, using examples to illustrate the decision-making process and the flexibility of decision trees in representing boolean functions. it also explains the basic steps of the id3 decision tree algorithm, including choosing the best decision attribute, creating descendants for different outcomes, and stopping when all training examples have the same class, with a focus on the decisions of when to continue and which attribute to use for the test.", 'chapters': [{'end': 1326.767, 'start': 1193.146, 'title': 'Decision trees for tennis', 'summary': "Discusses the concept of decision trees for determining whether it's a good day to play tennis, using examples to illustrate the decision-making process and the flexibility of decision trees in representing boolean functions.", 'duration': 133.621, 'highlights': ["Decision tree for determining whether it's a good day to play tennis based on attributes like outlook, humidity, wind, and temperature.", 'Illustration of the decision-making process using a sample decision tree and a new example, determining the outcome based on attribute values.', 'Flexibility of decision trees in representing Boolean functions as a disjunction of conjunctions, allowing representation of all Boolean functions.']}, {'end': 1460.321, 'start': 1326.767, 'title': 'Decision tree algorithm basics', 'summary': 'Explains the basic steps of the id3 decision tree algorithm, including choosing the best decision attribute, creating descendants for different outcomes, and stopping when all training examples have the same class, with a focus on the decisions of when to continue and which attribute to use for the test.', 'duration': 133.554, 'highlights': ['The algorithm is called top down induction of decision tree (ID3) and was proposed by Quinlan, focusing on choosing the best decision attribute at each node and splitting the training examples into different branches.', 'At a particular node, if all the training examples have the same class, the process can be stopped, highlighting the key decision point in the algorithm.', 'The chapter emphasizes the decisions involved in the algorithm, such as deciding when to continue, which attribute to use for the test, and when to stop, setting the stage for specific heuristics to be discussed in the next class.']}], 'duration': 267.175, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/FuJVLsZYkuE/pics/FuJVLsZYkuE1193146.jpg', 'highlights': ["Decision tree for determining whether it's a good day to play tennis based on attributes like outlook, humidity, wind, and temperature.", 'Illustration of the decision-making process using a sample decision tree and a new example, determining the outcome based on attribute values.', 'Flexibility of decision trees in representing Boolean functions as a disjunction of conjunctions, allowing representation of all Boolean functions.', 'The algorithm is called top down induction of decision tree (ID3) and was proposed by Quinlan, focusing on choosing the best decision attribute at each node and splitting the training examples into different branches.', 'At a particular node, if all the training examples have the same class, the process can be stopped, highlighting the key decision point in the algorithm.', 'The chapter emphasizes the decisions involved in the algorithm, such as deciding when to continue, which attribute to use for the test, and when to stop, setting the stage for specific heuristics to be discussed in the next class.']}], 'highlights': ['Decision trees are a non-linear function and a tree-structured classifier with decision and leaf nodes.', 'Decision trees can be used for both classification and regression, with a primary focus on classification.', 'The process of navigating a decision tree involves starting at the root and moving through the branches based on test values until reaching a leaf node.', 'The decision tree example consists of 3 decision nodes and 4 leaf nodes, demonstrating the process of determining loan approval based on employment status, credit score, and income.', 'Using decision trees to analyze applicant attributes for loan approval or rejection The training set includes attributes such as income, employment status, credit score, and others to determine loan approval or rejection.', 'Using decision trees to predict computer purchase based on age, student status, and credit rating The chapter explains the use of decision trees to predict computer purchase based on various attributes such as age, student status, and credit rating.', 'The process of using a decision tree to predict car mileage based on weight and horsepower is explained, determining high or low mileage based on specific criteria.', 'The challenge of selecting the best decision tree among multiple possibilities is discussed, considering the potential presence of noisy data and the need to minimize errors.', 'Attributes such as author, thread, length, and user location are used to determine whether a user reads or skips a thread, illustrating the practical application of decision trees for action prediction.', 'The preference for smaller decision trees is commonly expressed as a bias, aiming for trees with smaller depth or fewer nodes, which restricts the hypothesis space (Week 1).', 'Finding the smallest decision tree that fits the data is a computationally hard problem, leading to the exploration of greedy algorithms to search for a good tree in the space of decision trees.', 'The training examples are used to recursively build decision trees by choosing attributes for node splitting, with the aim of reaching a leaf quickly for a smaller tree.', 'The decision tree building process involves making decisions at each node, including whether to stop growing the tree and which attribute to split on, based on the examples and their attributes.', 'Attributes such as length and thread are used for node splitting, with specific examples of the number of examples for each attribute value and the decision-making process.', 'Examples of decision trees are provided, demonstrating different approaches to reaching a leaf node, either with specific values or with dominant values.', "Decision tree for determining whether it's a good day to play tennis based on attributes like outlook, humidity, wind, and temperature.", 'Illustration of the decision-making process using a sample decision tree and a new example, determining the outcome based on attribute values.', 'Flexibility of decision trees in representing Boolean functions as a disjunction of conjunctions, allowing representation of all Boolean functions.', 'The algorithm is called top down induction of decision tree (ID3) and was proposed by Quinlan, focusing on choosing the best decision attribute at each node and splitting the training examples into different branches.', 'At a particular node, if all the training examples have the same class, the process can be stopped, highlighting the key decision point in the algorithm.', 'The chapter emphasizes the decisions involved in the algorithm, such as deciding when to continue, which attribute to use for the test, and when to stop, setting the stage for specific heuristics to be discussed in the next class.']}