[{"data":1,"prerenderedAt":1163},["ShallowReactive",2],{"$J0WAj1kDXu":3},{"id":4,"title":5,"body":6,"description":82,"extension":1156,"meta":1157,"navigation":1158,"path":1159,"seo":1160,"stem":1161,"__hash__":1162},"content/modules/classification/logistic-regression/content.md","Logistic Regression",{"type":7,"value":8,"toc":1111},"minimark",[9,13,18,27,34,38,41,60,64,69,72,83,89,95,98,102,113,116,122,127,153,156,160,163,169,173,182,186,189,193,196,202,206,213,219,222,242,247,258,262,265,271,274,278,281,287,294,298,301,305,314,318,321,341,344,348,351,370,374,377,383,403,406,410,417,423,426,432,437,448,453,464,468,471,475,478,484,486,512,516,519,525,530,549,555,559,562,568,573,581,585,588,594,598,606,610,613,619,624,635,639,658,662,665,695,698,702,705,738,742,745,765,769,772,789,793,831,835,838,882,886,986,990,994,997,1011,1015,1035,1039,1042,1056,1059,1073,1077,1080],[10,11,5],"h1",{"id":12},"logistic-regression",[14,15,17],"h2",{"id":16},"introduction","Introduction",[19,20,21,22,26],"p",{},"Despite its name, logistic regression is a ",[23,24,25],"strong",{},"classification"," algorithm, not a regression algorithm. It's one of the most fundamental and widely-used algorithms for binary classification problems - situations where we need to predict one of two possible outcomes.",[19,28,29,30,33],{},"The key innovation of logistic regression is the ",[23,31,32],{},"sigmoid function",", which transforms a linear combination of features into a probability between 0 and 1. This makes it perfect for answering yes/no questions: Will a customer buy? Is this email spam? Will a patient develop a disease?",[14,35,37],{"id":36},"what-youll-learn","What You'll Learn",[19,39,40],{},"By the end of this module, you will:",[42,43,44,48,51,54,57],"ul",{},[45,46,47],"li",{},"Understand how the sigmoid function maps predictions to probabilities",[45,49,50],{},"Learn how binary cross-entropy loss guides the learning process",[45,52,53],{},"Interpret classification metrics (accuracy, precision, recall, F1 score)",[45,55,56],{},"Visualize decision boundaries that separate classes",[45,58,59],{},"Recognize the impact of learning rate and regularization on model performance",[14,61,63],{"id":62},"the-logistic-model","The Logistic Model",[65,66,68],"h3",{"id":67},"from-linear-to-logistic","From Linear to Logistic",[19,70,71],{},"Linear regression predicts continuous values:",[73,74,79],"pre",{"className":75,"code":77,"language":78},[76],"language-text","y = w₁x₁ + w₂x₂ + ... + wₙxₙ + w₀\n","text",[80,81,77],"code",{"__ignoreMap":82},"",[19,84,85,86,88],{},"But for classification, we need probabilities (values between 0 and 1). Logistic regression applies the ",[23,87,32],{}," to the linear combination:",[73,90,93],{"className":91,"code":92,"language":78},[76],"P(y=1|x) = σ(w₁x₁ + w₂x₂ + ... + wₙxₙ + w₀)\n",[80,94,92],{"__ignoreMap":82},[19,96,97],{},"Where σ is the sigmoid function.",[65,99,101],{"id":100},"the-sigmoid-function","The Sigmoid Function",[19,103,104,109],{},[105,106],"img",{"alt":107,"src":108},"Sigmoid Function","https://upload.wikimedia.org/wikipedia/commons/thumb/8/88/Logistic-curve.svg/500px-Logistic-curve.svg.png",[110,111,112],"em",{},"The sigmoid function maps any real number to a value between 0 and 1",[19,114,115],{},"The sigmoid function is defined as:",[73,117,120],{"className":118,"code":119,"language":78},[76],"σ(z) = 1 / (1 + e⁻ᶻ)\n",[80,121,119],{"__ignoreMap":82},[19,123,124],{},[23,125,126],{},"Key Properties:",[42,128,129,135,141,147],{},[45,130,131,134],{},[23,132,133],{},"Output range",": Always between 0 and 1",[45,136,137,140],{},[23,138,139],{},"S-shaped curve",": Smooth transition from 0 to 1",[45,142,143,146],{},[23,144,145],{},"Midpoint",": σ(0) = 0.5",[45,148,149,152],{},[23,150,151],{},"Asymptotes",": As z → ∞, σ(z) → 1; as z → -∞, σ(z) → 0",[19,154,155],{},"This makes it perfect for representing probabilities!",[65,157,159],{"id":158},"making-predictions","Making Predictions",[19,161,162],{},"Once we have the probability, we classify using a threshold (typically 0.5):",[73,164,167],{"className":165,"code":166,"language":78},[76],"Predicted class = {\n  1  if P(y=1|x) ≥ 0.5\n  0  if P(y=1|x) \u003C 0.5\n}\n",[80,168,166],{"__ignoreMap":82},[14,170,172],{"id":171},"how-it-works-gradient-descent-with-binary-cross-entropy","How It Works: Gradient Descent with Binary Cross-Entropy",[19,174,175,179],{},[105,176],{"alt":177,"src":178},"Logistic Regression Decision Boundary","https://upload.wikimedia.org/wikipedia/commons/thumb/6/6d/Exam_pass_logistic_curve.svg/500px-Exam_pass_logistic_curve.svg.png",[110,180,181],{},"Logistic regression creates a smooth decision boundary between classes",[65,183,185],{"id":184},"step-1-initialize-parameters","Step 1: Initialize Parameters",[19,187,188],{},"Start with random (or zero) values for all weights and the bias.",[65,190,192],{"id":191},"step-2-compute-probabilities","Step 2: Compute Probabilities",[19,194,195],{},"For each data point, calculate the probability of belonging to class 1:",[73,197,200],{"className":198,"code":199,"language":78},[76],"pᵢ = σ(w₁x₁ᵢ + w₂x₂ᵢ + ... + wₙxₙᵢ + w₀)\n",[80,201,199],{"__ignoreMap":82},[65,203,205],{"id":204},"step-3-calculate-loss","Step 3: Calculate Loss",[19,207,208,209,212],{},"Use ",[23,210,211],{},"binary cross-entropy loss"," to measure prediction quality:",[73,214,217],{"className":215,"code":216,"language":78},[76],"Loss = -(1/m) Σ[yᵢ·log(pᵢ) + (1-yᵢ)·log(1-pᵢ)]\n",[80,218,216],{"__ignoreMap":82},[19,220,221],{},"Where:",[42,223,224,230,236],{},[45,225,226,229],{},[23,227,228],{},"m"," is the number of data points",[45,231,232,235],{},[23,233,234],{},"yᵢ"," is the true label (0 or 1)",[45,237,238,241],{},[23,239,240],{},"pᵢ"," is the predicted probability",[19,243,244],{},[23,245,246],{},"Why this loss function?",[42,248,249,252,255],{},[45,250,251],{},"When yᵢ = 1: Loss = -log(pᵢ), so we want pᵢ close to 1",[45,253,254],{},"When yᵢ = 0: Loss = -log(1-pᵢ), so we want pᵢ close to 0",[45,256,257],{},"The logarithm heavily penalizes confident wrong predictions",[65,259,261],{"id":260},"step-4-compute-gradients","Step 4: Compute Gradients",[19,263,264],{},"Calculate how much each weight should change to reduce the loss. The gradient for logistic regression is:",[73,266,269],{"className":267,"code":268,"language":78},[76],"∂Loss/∂wⱼ = (1/m) Σ(pᵢ - yᵢ)·xⱼᵢ\n",[80,270,268],{"__ignoreMap":82},[19,272,273],{},"Notice this looks similar to linear regression, but pᵢ is now a probability from the sigmoid function!",[65,275,277],{"id":276},"step-5-update-parameters","Step 5: Update Parameters",[19,279,280],{},"Adjust the weights to reduce the loss:",[73,282,285],{"className":283,"code":284,"language":78},[76],"wⱼ = wⱼ - α × ∂Loss/∂wⱼ\nw₀ = w₀ - α × ∂Loss/∂w₀\n",[80,286,284],{"__ignoreMap":82},[19,288,289,290,293],{},"Where ",[23,291,292],{},"α"," is the learning rate.",[65,295,297],{"id":296},"step-6-repeat","Step 6: Repeat",[19,299,300],{},"Continue until convergence or for a specified number of epochs.",[14,302,304],{"id":303},"key-hyperparameters","Key Hyperparameters",[19,306,307,311],{},[105,308],{"alt":309,"src":310},"Binary Cross-Entropy Loss","https://upload.wikimedia.org/wikipedia/commons/thumb/8/8c/Logistic_cost_function.png/500px-Logistic_cost_function.png",[110,312,313],{},"Binary cross-entropy loss function for logistic regression",[65,315,317],{"id":316},"learning-rate-α","Learning Rate (α)",[19,319,320],{},"Controls how quickly the model learns:",[42,322,323,329,335],{},[45,324,325,328],{},[23,326,327],{},"Too small",": Slow convergence, many epochs needed",[45,330,331,334],{},[23,332,333],{},"Too large",": May overshoot and fail to converge",[45,336,337,340],{},[23,338,339],{},"Just right",": Efficient convergence to good solution",[19,342,343],{},"For logistic regression, typical values are often higher than linear regression (0.01 to 1.0) because the sigmoid function naturally bounds the outputs.",[65,345,347],{"id":346},"epochs","Epochs",[19,349,350],{},"Number of complete passes through the training data:",[42,352,353,359,365],{},[45,354,355,358],{},[23,356,357],{},"Too few",": Model hasn't learned the pattern (underfitting)",[45,360,361,364],{},[23,362,363],{},"Too many",": Wastes computation after convergence",[45,366,367,369],{},[23,368,339],{},": Loss and accuracy stabilize",[65,371,373],{"id":372},"regularization-λ","Regularization (λ)",[19,375,376],{},"Prevents overfitting by penalizing large weights:",[73,378,381],{"className":379,"code":380,"language":78},[76],"Loss = Binary Cross-Entropy + (λ/2m) Σwⱼ²\n",[80,382,380],{"__ignoreMap":82},[42,384,385,391,397],{},[45,386,387,390],{},[23,388,389],{},"λ = 0",": No regularization, may overfit",[45,392,393,396],{},[23,394,395],{},"Small λ",": Slight penalty on complexity",[45,398,399,402],{},[23,400,401],{},"Large λ",": Strong penalty, simpler model, may underfit",[19,404,405],{},"Common values: 0 to 1.0",[14,407,409],{"id":408},"decision-boundaries","Decision Boundaries",[19,411,412,413,416],{},"The ",[23,414,415],{},"decision boundary"," is where the model switches from predicting class 0 to class 1. It occurs where:",[73,418,421],{"className":419,"code":420,"language":78},[76],"P(y=1|x) = 0.5\n",[80,422,420],{"__ignoreMap":82},[19,424,425],{},"Which means:",[73,427,430],{"className":428,"code":429,"language":78},[76],"w₁x₁ + w₂x₂ + ... + wₙxₙ + w₀ = 0\n",[80,431,429],{"__ignoreMap":82},[19,433,434],{},[23,435,436],{},"For 2D data (two features):",[42,438,439,442,445],{},[45,440,441],{},"The decision boundary is a straight line",[45,443,444],{},"Points on one side are classified as class 0",[45,446,447],{},"Points on the other side are classified as class 1",[19,449,450],{},[23,451,452],{},"Properties:",[42,454,455,458,461],{},[45,456,457],{},"Linear decision boundary (straight line/plane/hyperplane)",[45,459,460],{},"Can't capture complex non-linear patterns without feature engineering",[45,462,463],{},"Position and orientation determined by learned weights",[14,465,467],{"id":466},"classification-metrics","Classification Metrics",[19,469,470],{},"Unlike regression, we can't use MSE or RMSE. Instead, we use classification-specific metrics:",[65,472,474],{"id":473},"confusion-matrix","Confusion Matrix",[19,476,477],{},"A table showing prediction outcomes:",[73,479,482],{"className":480,"code":481,"language":78},[76],"                Predicted\n                0       1\nActual  0      TN      FP\n        1      FN      TP\n",[80,483,481],{"__ignoreMap":82},[19,485,221],{},[42,487,488,494,500,506],{},[45,489,490,493],{},[23,491,492],{},"TP (True Positive)",": Correctly predicted class 1",[45,495,496,499],{},[23,497,498],{},"TN (True Negative)",": Correctly predicted class 0",[45,501,502,505],{},[23,503,504],{},"FP (False Positive)",": Incorrectly predicted class 1 (Type I error)",[45,507,508,511],{},[23,509,510],{},"FN (False Negative)",": Incorrectly predicted class 0 (Type II error)",[65,513,515],{"id":514},"accuracy","Accuracy",[19,517,518],{},"Proportion of correct predictions:",[73,520,523],{"className":521,"code":522,"language":78},[76],"Accuracy = (TP + TN) / (TP + TN + FP + FN)\n",[80,524,522],{"__ignoreMap":82},[19,526,527],{},[23,528,529],{},"Interpretation:",[42,531,532,538,543],{},[45,533,534,537],{},[23,535,536],{},"Range",": 0 to 1 (or 0% to 100%)",[45,539,540],{},[23,541,542],{},"Higher is better",[45,544,545,548],{},[23,546,547],{},"Limitation",": Can be misleading with imbalanced classes",[19,550,551,554],{},[23,552,553],{},"Example:"," If 95% of emails are not spam, a model that always predicts \"not spam\" has 95% accuracy but is useless!",[65,556,558],{"id":557},"precision","Precision",[19,560,561],{},"Of all positive predictions, how many were actually positive?",[73,563,566],{"className":564,"code":565,"language":78},[76],"Precision = TP / (TP + FP)\n",[80,567,565],{"__ignoreMap":82},[19,569,570],{},[23,571,572],{},"When to prioritize:",[42,574,575,578],{},[45,576,577],{},"When false positives are costly",[45,579,580],{},"Example: Medical diagnosis - don't want to tell healthy people they're sick",[65,582,584],{"id":583},"recall-sensitivity","Recall (Sensitivity)",[19,586,587],{},"Of all actual positives, how many did we correctly identify?",[73,589,592],{"className":590,"code":591,"language":78},[76],"Recall = TP / (TP + FN)\n",[80,593,591],{"__ignoreMap":82},[19,595,596],{},[23,597,572],{},[42,599,600,603],{},[45,601,602],{},"When false negatives are costly",[45,604,605],{},"Example: Disease screening - don't want to miss sick people",[65,607,609],{"id":608},"f1-score","F1 Score",[19,611,612],{},"Harmonic mean of precision and recall:",[73,614,617],{"className":615,"code":616,"language":78},[76],"F1 = 2 × (Precision × Recall) / (Precision + Recall)\n",[80,618,616],{"__ignoreMap":82},[19,620,621],{},[23,622,623],{},"Why harmonic mean?",[42,625,626,629,632],{},[45,627,628],{},"Balances precision and recall",[45,630,631],{},"Penalizes extreme values",[45,633,634],{},"Only high if both precision and recall are high",[19,636,637],{},[23,638,529],{},[42,640,641,646,652],{},[45,642,643,645],{},[23,644,536],{},": 0 to 1",[45,647,648,651],{},[23,649,650],{},"1.0",": Perfect precision and recall",[45,653,654,657],{},[23,655,656],{},"Low",": Poor performance on at least one metric",[14,659,661],{"id":660},"the-precision-recall-tradeoff","The Precision-Recall Tradeoff",[19,663,664],{},"There's often a tradeoff between precision and recall:",[42,666,667,681],{},[45,668,669,672,673],{},[23,670,671],{},"High threshold (e.g., 0.8)",": Predict class 1 only when very confident",[42,674,675,678],{},[45,676,677],{},"Higher precision (fewer false positives)",[45,679,680],{},"Lower recall (more false negatives)",[45,682,683,686,687],{},[23,684,685],{},"Low threshold (e.g., 0.3)",": Predict class 1 more liberally",[42,688,689,692],{},[45,690,691],{},"Lower precision (more false positives)",[45,693,694],{},"Higher recall (fewer false negatives)",[19,696,697],{},"The F1 score helps balance these competing objectives.",[14,699,701],{"id":700},"assumptions-and-requirements","Assumptions and Requirements",[19,703,704],{},"For logistic regression to work well:",[706,707,708,714,720,726,732],"ol",{},[45,709,710,713],{},[23,711,712],{},"Binary outcome",": Target must be 0 or 1 (can be extended to multi-class)",[45,715,716,719],{},[23,717,718],{},"Independent observations",": Data points should be independent",[45,721,722,725],{},[23,723,724],{},"Linear decision boundary",": Classes should be approximately linearly separable",[45,727,728,731],{},[23,729,730],{},"No multicollinearity",": Features shouldn't be highly correlated",[45,733,734,737],{},[23,735,736],{},"Large sample size",": More data generally improves performance",[14,739,741],{"id":740},"when-to-use-logistic-regression","When to Use Logistic Regression",[19,743,744],{},"Logistic regression is ideal when:",[42,746,747,750,753,756,759,762],{},[45,748,749],{},"You have a binary classification problem",[45,751,752],{},"You need an interpretable model",[45,754,755],{},"You want to understand feature importance",[45,757,758],{},"Classes are approximately linearly separable",[45,760,761],{},"You need probability estimates, not just class labels",[45,763,764],{},"You want fast training and prediction",[14,766,768],{"id":767},"limitations","Limitations",[19,770,771],{},"Logistic regression may not work well when:",[42,773,774,777,780,783,786],{},[45,775,776],{},"Classes are not linearly separable",[45,778,779],{},"There are complex non-linear relationships",[45,781,782],{},"You have many irrelevant features",[45,784,785],{},"Classes are highly imbalanced (without adjustments)",[45,787,788],{},"You need to capture feature interactions (without engineering them)",[14,790,792],{"id":791},"tips-for-better-results","Tips for Better Results",[706,794,795,801,807,813,819,825],{},[45,796,797,800],{},[23,798,799],{},"Feature Scaling",": Normalize or standardize features for faster convergence",[45,802,803,806],{},[23,804,805],{},"Feature Engineering",": Create polynomial or interaction features for non-linear boundaries",[45,808,809,812],{},[23,810,811],{},"Handle Imbalance",": Use class weights or resampling for imbalanced datasets",[45,814,815,818],{},[23,816,817],{},"Regularization",": Use L2 regularization to prevent overfitting",[45,820,821,824],{},[23,822,823],{},"Threshold Tuning",": Adjust the classification threshold based on your precision/recall needs",[45,826,827,830],{},[23,828,829],{},"Cross-Validation",": Validate performance on unseen data",[14,832,834],{"id":833},"real-world-applications","Real-World Applications",[19,836,837],{},"Logistic regression is used extensively in:",[42,839,840,846,852,858,864,870,876],{},[45,841,842,845],{},[23,843,844],{},"Healthcare",": Disease diagnosis, patient risk assessment",[45,847,848,851],{},[23,849,850],{},"Finance",": Credit scoring, fraud detection, default prediction",[45,853,854,857],{},[23,855,856],{},"Marketing",": Customer churn prediction, conversion optimization",[45,859,860,863],{},[23,861,862],{},"Email",": Spam detection",[45,865,866,869],{},[23,867,868],{},"E-commerce",": Purchase prediction, recommendation systems",[45,871,872,875],{},[23,873,874],{},"HR",": Employee attrition prediction",[45,877,878,881],{},[23,879,880],{},"Insurance",": Claim likelihood, risk assessment",[14,883,885],{"id":884},"comparison-with-linear-regression","Comparison with Linear Regression",[887,888,889,904],"table",{},[890,891,892],"thead",{},[893,894,895,899,902],"tr",{},[896,897,898],"th",{},"Aspect",[896,900,901],{},"Linear Regression",[896,903,5],{},[905,906,907,921,934,947,960,973],"tbody",{},[893,908,909,915,918],{},[910,911,912],"td",{},[23,913,914],{},"Task",[910,916,917],{},"Regression (continuous)",[910,919,920],{},"Classification (binary)",[893,922,923,928,931],{},[910,924,925],{},[23,926,927],{},"Output",[910,929,930],{},"Any real number",[910,932,933],{},"Probability (0 to 1)",[893,935,936,941,944],{},[910,937,938],{},[23,939,940],{},"Activation",[910,942,943],{},"None (identity)",[910,945,946],{},"Sigmoid",[893,948,949,954,957],{},[910,950,951],{},[23,952,953],{},"Loss Function",[910,955,956],{},"Mean Squared Error",[910,958,959],{},"Binary Cross-Entropy",[893,961,962,967,970],{},[910,963,964],{},[23,965,966],{},"Metrics",[910,968,969],{},"MSE, RMSE, R²",[910,971,972],{},"Accuracy, Precision, Recall, F1",[893,974,975,980,983],{},[910,976,977],{},[23,978,979],{},"Decision Boundary",[910,981,982],{},"N/A",[910,984,985],{},"Linear hyperplane",[14,987,989],{"id":988},"extensions","Extensions",[65,991,993],{"id":992},"multi-class-classification","Multi-class Classification",[19,995,996],{},"Logistic regression can be extended to multiple classes using:",[42,998,999,1005],{},[45,1000,1001,1004],{},[23,1002,1003],{},"One-vs-Rest (OvR)",": Train one classifier per class",[45,1006,1007,1010],{},[23,1008,1009],{},"Softmax Regression",": Generalization using softmax function",[65,1012,1014],{"id":1013},"regularization-variants","Regularization Variants",[42,1016,1017,1023,1029],{},[45,1018,1019,1022],{},[23,1020,1021],{},"Ridge (L2)",": Penalizes sum of squared weights",[45,1024,1025,1028],{},[23,1026,1027],{},"Lasso (L1)",": Penalizes sum of absolute weights, can zero out features",[45,1030,1031,1034],{},[23,1032,1033],{},"Elastic Net",": Combination of L1 and L2",[14,1036,1038],{"id":1037},"summary","Summary",[19,1040,1041],{},"Logistic regression is a powerful classification algorithm that:",[42,1043,1044,1047,1050,1053],{},[45,1045,1046],{},"Uses the sigmoid function to predict probabilities",[45,1048,1049],{},"Optimizes binary cross-entropy loss with gradient descent",[45,1051,1052],{},"Provides interpretable linear decision boundaries",[45,1054,1055],{},"Serves as a foundation for neural networks (it's essentially a single-layer neural network!)",[19,1057,1058],{},"Understanding logistic regression is crucial because:",[42,1060,1061,1064,1067,1070],{},[45,1062,1063],{},"It introduces key classification concepts",[45,1065,1066],{},"It's the building block for neural networks",[45,1068,1069],{},"It's widely used in industry for interpretable classification",[45,1071,1072],{},"The concepts extend to more complex algorithms",[14,1074,1076],{"id":1075},"next-steps","Next Steps",[19,1078,1079],{},"After mastering logistic regression, you can explore:",[42,1081,1082,1087,1093,1099,1105],{},[45,1083,1084,1086],{},[23,1085,993],{},": Softmax regression for more than two classes",[45,1088,1089,1092],{},[23,1090,1091],{},"Support Vector Machines (SVM)",": More sophisticated decision boundaries",[45,1094,1095,1098],{},[23,1096,1097],{},"Decision Trees",": Non-linear, interpretable classification",[45,1100,1101,1104],{},[23,1102,1103],{},"Neural Networks",": Stacking multiple logistic regression units",[45,1106,1107,1110],{},[23,1108,1109],{},"Ensemble Methods",": Combining multiple classifiers for better performance",{"title":82,"searchDepth":1112,"depth":1113,"links":1114},2,3,[1115,1116,1117,1122,1130,1135,1136,1143,1144,1145,1146,1147,1148,1149,1150,1154,1155],{"id":16,"depth":1112,"text":17},{"id":36,"depth":1112,"text":37},{"id":62,"depth":1112,"text":63,"children":1118},[1119,1120,1121],{"id":67,"depth":1113,"text":68},{"id":100,"depth":1113,"text":101},{"id":158,"depth":1113,"text":159},{"id":171,"depth":1112,"text":172,"children":1123},[1124,1125,1126,1127,1128,1129],{"id":184,"depth":1113,"text":185},{"id":191,"depth":1113,"text":192},{"id":204,"depth":1113,"text":205},{"id":260,"depth":1113,"text":261},{"id":276,"depth":1113,"text":277},{"id":296,"depth":1113,"text":297},{"id":303,"depth":1112,"text":304,"children":1131},[1132,1133,1134],{"id":316,"depth":1113,"text":317},{"id":346,"depth":1113,"text":347},{"id":372,"depth":1113,"text":373},{"id":408,"depth":1112,"text":409},{"id":466,"depth":1112,"text":467,"children":1137},[1138,1139,1140,1141,1142],{"id":473,"depth":1113,"text":474},{"id":514,"depth":1113,"text":515},{"id":557,"depth":1113,"text":558},{"id":583,"depth":1113,"text":584},{"id":608,"depth":1113,"text":609},{"id":660,"depth":1112,"text":661},{"id":700,"depth":1112,"text":701},{"id":740,"depth":1112,"text":741},{"id":767,"depth":1112,"text":768},{"id":791,"depth":1112,"text":792},{"id":833,"depth":1112,"text":834},{"id":884,"depth":1112,"text":885},{"id":988,"depth":1112,"text":989,"children":1151},[1152,1153],{"id":992,"depth":1113,"text":993},{"id":1013,"depth":1113,"text":1014},{"id":1037,"depth":1112,"text":1038},{"id":1075,"depth":1112,"text":1076},"md",{},true,"/modules/classification/logistic-regression/content",{"title":5,"description":82},"modules/classification/logistic-regression/content","wy-KqSfB732V__X7XPjSQf-3eovKZCbTQDcswen4z3M",1792223550729]