1. A startup is collecting IoT data from sensors placed on manufacturing equipment. The sensors send data every five seconds. The data includes a machine identifier, a timestamp, and several numeric values. The startup is developing a model to identify unusual readings. What type of unsupervised learning technique would they use?

2. You want to study deep learning and decide to start with the basics. You build a binary classifier using an artificial neuron. What algorithm would you use to train it?

3. You are building a machine learning model to predict the sales price of houses. You have 7 years of historical data, including 18 features of houses and their sales price. What type of machine learning algorithm would you use?

4. You have been tasked with developing a classification model. You have reviewed the data that you will use for training and testing and realize that there are a number of outliers that you think might lead to overfitting. What technique would you use to reduce the impact of those outliers on the model?

5. You are preparing a dataset to build a classifier. The data includes several continuous values, each in the range 0.00 to 100.00. You’d like to have a discrete feature derive each continuous value. What type of feature engineering would you use?

6. You are reviewing a dataset and find that the data is relatively high quality. There are no missing values and only a few outliers. You build a model based on the dataset that has high accuracy, precision, and recall when applied to the test data. When you use the model in production, however, it renders poor results. What might have caused this condition?

7. You have been asked to build a machine learning model that will predict if a news article is a story about technology or another topic. Which of the following would you use?

8. Your team is building a classifier to identify counterfeit products on an e-commerce site. Most of the products on the site are legitimate, and only about 3 percent of the products are counterfeit. You are concerned that, as is, the dataset will lead to a model that always predicts that products are legitimate. Which of the following techniques could you use to prevent this?

9. You have built a deep learning neural network that has 8 layers, and each layer has 100 fully connected nodes. The model fits the training data quite well with an F1 score of 98 out of 100. The model performs poorly when the test data is used, resulting in an F1 score of 62 out of 100. What technique would you use to try to improve performance of this model?

10. A group of machine learning engineers has been assigned the task of building a machine learning model to predict the price of gold on the open market. Many features could be used, and the engineers believe that the optimal model will be complex. They want to understand the minimum predictive value of a model that they can build from the data that they have. What would they build?