domain 3 AI gmetrix

1) The systematic pattern of deviation from the norm in judgment. 2)Lack of context around data. 3)Supports prior beliefs or values. 4)Result of cultural stereotypes in the people involved.

1) Cognitive bias 2)Analytics bias 3)Confirmation bias 4)Prejudice bias

1)Dichotomizes the result of a quantitative test 2)Optimal decision threshold 3)Thresholds factoring in the costs of misdiagnosis 4)Tests if bias meets performance requirements

1) Decision Threshold 2)Diagnostic Accuracy 3)Best ROC Curve 4)Deming Regression

1) Estimate next year's sales based on the sales of the current year 2) Optimize the behavior of an agent-based on feedback from the environment 3) Detect anomalies based on data's internal patterns 4) Provide autocorrect, autocomplete, and predictive text on smartphones

1) Forecasting Supervised Learning Model 2)Reinforcement Learning Model 3) Unsupervised Learning Models 4)Natural Language Processing Models

Step 1 Step 2 Step 3 Step 4

1) Initialize weights of perceptron randomly or based on an algorithm. 2)For a sample input, compute an output. 3)If the prediction does not match the output, change the weights. 4) Go to the next batch of the dataset

1)ROC CURVE 2)Area Under the Curve (AUC) 3)F1-Score 4)Mean Squared Error (MSE)

1) It measures the True Positive Rate (TPR) against the False Positive Rate (FPR). 2) It measures the ability of a classifier to distinguish between classes. 3)It measures the model's accuracy as the harmonic mean of precision and recall. 4)It measures regression prediction performance.

1) The measure of the degree of separation between the positive and negative distributions 2) The range of potential values of an unknown population parameter 3) Randomly split the dataset into training and testing partitions 4) Split the dataset into k-partitions or folds

1) KSC 2) Confidence Interval 3)Percentage Split 4) Cross-Validation

1)Visualize the effect of the change for a certain feature globally 2)Visualize the effect of the change for a certain feature locally 3)Black-box machine learning predictions 4)Explain the output of any machine learning model using a game theory approach

1) Partial dependence plots 2)Individual conditional expectation plots 3)Global surrogate models 4) Shapley Additive Explanations

1) Sensitivity/(1-Specificity) 2) (1- Sensitivity)/Specificity 3) 0.961 4) 0.906

1) Positive Likelihood Ratio 2)Negative Likelihood Ratio 3)Sensitivity 4)Specificity

Which two evaluation metrics are used for classification algorithms? (Choose 2.)

1)Area under the ROC Curve 2)F1-Score

1) Tests relationships between categorical variables 2)Measures the dependence of two variables 3)Measures the strength of the ordinal association between two variables 4)Measures the degree of association between two variables

1)Chi-squared Score 2)Analysis Of Variance Test 3)Kendall's Rank Correlation Coefficient 4)Spearman's Rank Correlation Coefficient

Which three methods are used to prevent overfitting? (Choose 3.)

1)Data simplification 2)Training with more data 3)Data augmentation

Which three questions must be considered when documenting a machine learning model? (Choose 3.)

1)How long does it take to generate predictions using the model? 2)Should it have a README.md file at the top level of the archive? 3)How should features be selected?

Which four aspects are useful when evaluating a machine learning solution to ensure the right problem got solved? (Choose 4.)

1)Key performance indicators 2)Lifetime and use 3)Rationale 4)Benefits

Select the two characteristics of model hyperparameters. (Choose 2.)

1)They are often specified by the practitioner. 2)They can often be set using heuristics.

Why are ensemble methods which combine several decision trees better than a single decision tree?

An ensemble method can make overall better predictions.

Select the correct option that represents a negative correlation in the answer area.

Area 3, which is the image on the right, is correct because it indicates a negative correlation.

Which family of machine learning algorithms can be used to detect cancer?

Classification algorithms

To which family of algorithms does k-means belong?

Clustering algorithms

Which algorithm is used for image classification?

Convolutional Neural Networks

Which type of bias listed below exists in online recruitment tools?

Gender bias

Given below is a scenario for training error, TE, and validation error, VE, for a machine learning algorithm. What is the best hyperparameter H based on TE and VE?

H- 2 TE-200 VE-73

Consider a training dataset used to predict a certain type of cancer. The dataset consists of 50% male patients and 50% female patients. 87% of patients have the normal gene and 13% of patients have oncogene. How could this dataset be defined based on bias?

Imbalanced Dataset

You are considering a training dataset used to predict cancer. The dataset consists of 50% male patients and 50% female patients. 90% of patients have the normal gene and 10% of patients have oncogene. How could this dataset be defined based on bias?

Imbalanced Dataset

Which answer option is an evaluation metric for regression algorithms?

Mean Absolute Error

While building a machine learning model to predict the price of a home, a team has been provided with previous years' datasets which include the location, housing square footage, etc. with values in number format and the housing rate (output) in number format. Which model would be best suited to predict the home price?

Multiple Linear Regression

Instead of random weight initializations in a neural network for a classification problem, all the weights are set to zero. Which statement below is correct?

Neurons will learn the same features in each iteration.

Which is a technique to convert an imbalanced dataset into a balanced dataset?

Oversampling

A corporation used two alternative ways to perform a classification task. A convolutional neural network (CNN) is used in one, while a recurrent neural network is used in the other (RNN). Both models have a comparable level of precision. What can be the most significant disadvantage of using RNN instead of CNN for this task?

RNNs are slower than CNNs in the matter of speed.

When detecting cancer, which of these metrics must be optimized above all others?

Recall

What does the learning curve below indicate?

The model is overfitting.

A logistic regression model with a defined classification threshold of 0.5 helps determine if an email is spam (positive) or not spam (negative). For a particular input email, the model generates a 0.4 probability score. What can be said about the input email?

The model will predict the input email is not spam.

An AI model is being built with a medical imagery dataset of malignant and benign tumors to predict if a patient will test positive or negative for cancer. Which combination of factors makes the most accurate model?

True Positive (48%), False Positive (85%), False Negative (46%), True Negative (98%)