Coursera Cerficate Course_1200_630

Managing Machine Learning Projects with Google Cloud 2023

by

Non-technical business personnel have a rare chance to direct or shape machine learning initiatives. This course is for you if you want to learn how to utilise machine learning without all the technical lingo and have questions about it. Learn how to convert business issues into use cases for machine learning and assess their viability and impact. Learn how to identify the phases of an ML project and the factors that go into each one, as well as how to find unexpected use cases. Gain the confidence you need to present a unique ML use case to your team or leadership or translate the needs to a technical team.

Enroll Here: Managing Machine Learning Projects with Google Cloud

Identifying business value for using ML Coursera Quiz/Assessment Answers

Question 1: Which description best defines machine learning?

  • Machine learning refers to a computer that can think and act like a human.
  • Machine learning refers to the process of fixing and cleaning data for analytics purposes.
  • Machine learning refers to the process of a computer carrying out many steps to complete a complex task.
  • Machine learning refers to a branch of AI and one way to solve a problem autonomously.

Question 2: You own a new local private transport service and want to use machine learning to adjust the pricing of your service throughout the day instead of using a fixed-pricing approach that does not change automatically. What data would a machine learning model need in order to predict when to adjust transport pricing?

  • The time of day and customer demand
  • The current level of traffic in the area
  • The length and route of a passenger’s journey
  • The average number of completed journeys within an hour

Question 3: Which phase of a machine learning project occurs before collecting and preparing data?

  • Assess whether a problem is right for machine learning.
  • Evaluate the previous machine learning model.
  • Set the metrics and objectives for a machine learning model.
  • Determine the time and resources needed to train a machine learning model.

Question 4: A franchise movie-rental store has decided to launch their own streaming rental service to expand their presence in the online market. The stores are well known for their in-depth and personalized recommendations by the store clerks. The CEO wants the new rental streaming service to also provide relevant recommendations for its customers. What data could be used to provide the most accurate customer recommendations in an online film and movie streaming service?

  • A customer’s genre preferences and previous purchases
  • A customer’s current age, location, and hobbies
  • The films with the highest overall watch time in each genre
  • The film that the store clerks would generally recommend in each genre

Question 5: You work for a graphic design software company that offers a 7-day trial of your products to potential paying customers. You analyze why users convert their free trial into a software purchase. You are curious whether machine learning can predict which customers will convert their trial into a purchase and which users might need additional persuasion. Assess the following use case against a two-factor grid:“How could we use machine learning to predict whether a trial user will convert to a paid purchase after using the software’s unique color editing tools?

  • This use case is specific but impossible.
  • This use case is ambiguous and impossible.
  • This use case is ambiguous and challenging.
  • This use case is clear and challenging.

Defining ML as a practice Coursera Quiz/Assessment Answers

Question 1: You lead a team of researchers within a bioinformatics organization that aims to automatically annotate the functions of proteins. You can add this output data to your existing knowledge database, which contains information about other protein annotations. What type of machine learning use case is this?

  • This is a classification use case.
  • This is a clustering use case.
  • This is a regression use case.
  • This use case requires both classification and regression.

Question 2: You own a model trained to predict life expectancy that was trained on data collected from Norwegians. This model is successful in the Norwegian region; however, when used in nearby Lithuania, performance declines considerably. What is most likely the source of this issue?

  • Data coverage
  • Data cleanliness
  • Data localization
  • Data consistency

Question 3: You work at a toy manufacturer that assembles action figures aimed at young children. Due to manufacturing faults, unsafe parts regularly need to be removed from the assembly line. How can machine learning scale the business intelligence needed in this quality control situation?

  • Machine learning analyzes which action figures are more prone to faults.
  • Machine learning makes repeated decisions on which parts to remove.
  • Machine learning predicts when a faulty part will appear on the assembly line.
  • Machine learning creates data for assembling different action figures in the future.

Question 4: At a national energy provider company, the data analytics team has been suggesting long-term energy pricing based on customer usage reports from each region. You are the new manager of the data analytics team and want to introduce machine learning into the analytics process for repeatability and scalability. What effect can predictive insights using machine learning have on energy pricing in this use case?

  • Machine learning can predict future weather events that significantly affect pricing.
  • Machine learning can help produce more accurate analytics reports of customer usage.
  • Machine learning can undercut the pricing of competitive energy providers across regions.
  • Machine learning can determine and adjust pricing in different regions instantly.

Question 5: Machine learning has existed since the 1970s. However, there has been a surge in usage across all industries only recently. As a manager, why might you choose to introduce machine learning processes into your business now?

  • Machine learning algorithms are becoming more sophisticated and mature.
  • The cost of machine learning projects is now affordable for most companies.
  • Companies can outsource data collection to ensure that there are no errors in their data.
  • Machine learning algorithms can be trained by large datasets within seconds using little computing power.

Building and evaluating ML models Coursera Quiz/Assessment Answers

Question 1: Input data in a machine learning model is often made of three parts, which are: the features of an example, the resulting label, and the label type. How can the ‘features’ be defined in a machine learning context?

  • The characteristics that give meaning to a piece of data
  • The classification of a piece of data in a category
  • The attributes that optimize a machine learning model
  • The patterns that a machine learning model labels after deployment

Question 2: You work at a medical research facility that analyzes patient data for local hospitals. You want to use machine learning for specialized image recognition in order to identify bacterial infections in patients’ x-ray images. What is the preferred method of obtaining a labeled dataset for this custom image recognition use case?

  • Use AutoML Vision to classify the x-ray images.
  • Use the Vision API to classify the x-ray images.
  • Use the REST API to classify the x-ray images.
  • Send the images to a third-party labeling company to classify the x-rays.

Question 3: You work at a mobile phone manufacturer and are preparing to launch the newest version of your high-end phone. You want to analyze the battery efficiency of your new phone against previous models. You have a backlog of historical data on previous models and their results, but these datasets exist in silos separate from the data for your new phone. How can you acquire a labeled dataset in this scenario when datasets exist in separate silos?

  • Use data on standby time as a substitute proxy dataset.
  • Choose one dataset to train your machine learning model on only.
  • Use a data warehouse to join the datasets into one source.
  • Train a machine learning model that produces more accessible data, and use the output data as your input data.

Question 4: You work at a car manufacturing company that is ready to deploy a machine learning model. However, you want to evaluate the model first and decide to evaluate your model with a small set of data. You cannot measure how accurate the model is on all the original training data because it could memorize all answers and perform badly after deployment. What is a reasonable percentage of the data to reserve when you are evaluating the accuracy of a machine learning model?

  • 10–25%
  • 1–5%
  • 5–10%
  • 25–40%

Question 5: You are the communications manager at a marketing company. Recently, you noticed an increase in spam marketing emails disguised as popular brand emails that you want to filter out of your inbox. You want to use machine learning to predict which emails are spam and should be filtered. What are some possible features in this machine learning use case to detect deceptive spam emails?

  • Email length and word count
  • Email logo usage and color accuracy
  • Email subject line and number of links
  • Email language and grammatical accuracy

Question 6: You lead the marketing team for a startup accomodation booking website. You want to provide users with personalized accommodation recommendations, but lack sufficient historical labeled data of customer bookings to use as an exclusive data source. Instead, you and your team have only been using user clicks and accommodation viewings as a proxy for your entire dataset. What is the issue of only using user clicks and accommodation viewings as your dataset that might lead to few converted bookings?

  • Customers might browse accommodations while having no intention of booking them.
  • Recommendations should be based on the customer’s previous bookings only.
  • A customer might view different accommodation pages for different lengths of time.
  • The website might make recommendations in locations the customer has already visited.

Question 7: You are a doctor at a small medical clinic studying the symptoms and effects of common health conditions. You want to use machine learning to predict which of your patients might have an increased probability of heart disease. However, you have a limited dataset due to having fewer patients than a full-sized hospital. What would be the preferred solution to identify patients with an increased probability of heart disease using machine learning?

  • Use existing data from a large nearby hospital as proxy data.
  • Use existing data of other health conditions as proxy data.
  • Join your data with the data of another doctor in your clinic.
  • Use existing data from a hospital that has high diagnosis rates of heart disease in patients.

Question 8: You are working on the data team at a global banking company. You are gathering a wide variety of labeled data from different departments and locations for future machine learning experiments. Before you can introduce the data to train the machine learning model, what do you need to do?

  • Prepare the data by categorizing it by location.
  • Remove any data irrelevant to the first machine learning project.
  • Prepare the data and store it in a single location.
  • Delete 80% of the data because it is not needed for training the model.

Question 9: You work in the customer retention team at a bank and have noticed an increase in customers leaving your service. To solve this problem, you use machine learning with an objective to improve customer retention at your bank by personalizing services and loans. What is the preferred optimization of your objective to improve customer experience and retention at your bank?

  • Provide offers based on customer spending behavior.
  • Provide offers that are the most popular services and loans.
  • Provide offers dependent on how many accounts a customer has.
  • Provide offers based on product advertisements a customer has viewed.

Question 10: Machine learning projects consist of many different phases. However, a lot of useful information cannot be described in the phases alone, such as guidance on machine learning best practices. What is considered an example of good practice in machine learning?

  • Testing your machine learning projects with end users
  • Carrying out machine learning tasks using R only
  • Using test data during machine learning experiments
  • Using one algorithm for all machine learning problems

Using ML responsibly and ethically Coursera Quiz/Assessment Answers

Question 1: Google has developed a set of AI principles that govern its research and product development and affect its business decisions. One of the AI principles is to avoid creating or reinforcing unfair bias. Why does a company need to adhere to a set of agreed-upon standards when working with AI?

  • How AI is developed and used will affect society.
  • To limit experimentation and use cases with AI.
  • To avoid an AI gaining too much knowledge of various subjects.
  • So that AI development and usage is not held accountable by government law.

Question 2: Your company has decided to experiment with artificial intelligence to improve business processes and decision-making efficiency. Your company has decided to follow the AI principles developed by Google, and one of the AI principles is to avoid creating or reinforcing unfair bias. Why is it important to avoid creating unfair biases in machine learning models?

  • Machine learning algorithms can reflect and reinforce biases that can have unjust affects on people.
  • Machine learning algorithms will be less accurate but more fair with unbiased data.
  • Machine learning algorithms cannot improve business processes and decision-making efficiency with biased data.
  • Machine learning algorithms do not show any undesirable behaviors, so biases cannot be found and resolved easily.

Question 3: You are a scientist at a hospital studying the likelihood of common health conditions in your community. You want to use machine learning to predict which patients might have an increased probability of heart disease. You already have decided to use recently updated patient data including: age range, gender, medical history, weight, and how often, if ever, a person smokes. In this scenario, where and why might you expect to find reporting bias?

  • A patient could lie about how often they smoke.
  • A patient might not disclose their true age.
  • A patient might have lied about their weight by a few pounds.
  • The medical history completed by the hospital might contain unknown inaccuracies.

Question 4: An international airport has tasked a technology company to create a machine learning model that can identify potential criminals entering the country. The airport has provided the company with images of criminal and non-criminal residents, with their consent. The airport has long queues at immigration and hopes that introducing this electronic system will improve efficiency and the traveler experience. In this scenario, what assumption is being made that is an example of automation bias?

  • The electronic system will decrease the queue time at immigration control.
  • The machine learning model will find a correlation between the criminals’ images that signifies potential criminal behavior.
  • The machine learning model will already be aware of known criminals, which flags them for potential criminal behavior.
  • A technology company with machine learning experience can make a security system that identifies potential criminals.

Question 5: Google has developed a set of AI principles that govern its research and product development and impact its business decisions. One of the AI principles is to avoid creating or reinforcing unfair biases. However, what should AI applications not do according to Google’s AI principles?

  • Breach human rights
  • Surpass human intelligence
  • Be used for recreational purposes
  • Produce indistinguishable output or results

Question 6: An owner of a national live sports TV channel has had ongoing success for 15+ years. Recently, the viewership on the channel has decreased due to viewers moving to online streaming platforms. The owner finds data indicating social media discussion for the sporting events on the channel is increasing every year, and argues that viewers are very much still engaged. The owner uses this data to justify the channel to advertisers who can buy advertisement time across the day. In this scenario, what is an indisputable example of confirmation bias?

  • The owner of the channel has had success for 15+ years and thinks this cannot change.
  • The owner used online streaming platforms as justification for decreasing viewership on the channel.
  • The owner believes that viewership of live sports TV channels cannot be affected by online streaming platforms.
  • The owner used social media discussion data as justification for viewer interest in the channel.

Question 7: You are a data collector at a movie review aggregation company that is aimed toward family and children’s movies. As audience members are leaving the screening room at the cinema, you survey their opinion of the movie on an alphabetical grading scale. Your company only surveys audiences within the first week of a movie’s release and then locks the grades. The aggregate grade is provided to film production companies who want to improve future family and children’s films. In this scenario, what would be an example of selection bias?

  • Only film production companies receive the final aggregated grade.
  • Only the reviews of family and children’s movies are being collected.
  • Only the opinions of the initial audiences at a cinema are included in the grade.
  • Only the grades of children, and not adults, are included in the aggregated grade.

Question 8: You are a translation vendor manager who has been tasked by a publishing company to translate historical Finnish texts. The translation is from Finnish, which has no gender-distinctive pronouns, to English. To save time, your experienced Finnish team decides to use a popular machine learning translation tool for longer texts while they translate the shorter texts manually. In this scenario, where might you expect an increased amplification of biases?

  • The translation tool will amplify biases through its gender pronoun assumptions in English.
  • The translation tool will amplify biases by using one gender pronoun throughout entire texts.
  • The vendor team will amplify biases through their own gender pronoun assumptions in English.
  • The vendor team will amplify biases by using one gender pronoun throughout entire texts.

Question 9: A survey is delivered over the internet through links within ads that appear on the homepage of popular newspapers’ websites. The survey asks sensitive questions about readers’ health and medical history. What type of bias is caused by not stating that surveyees might only include people who read the front page and not those who might consume primarily sports-related news?

  • Selection bias
  • Automation bias
  • Confirmation bias
  • Reporting bias

Question 10: A survey is delivered over the internet through links within ads that appear on the homepage of popular newspapers’ websites. The survey asks sensitive questions about readers’ health and medical history. It can be argued that selection bias exists in this scenario. What justification is there for this reasoning?

  • The ads might not be seen by older people who typically read physical newspapers.
  • The survey is performed exclusively online because it is easier.
  • The ads might only be seen by people who view the homepage of the site.
  • The survey will provide lackluster responses because only young people read online newspapers.

Discovering ML use cases in day-to-day business Coursera Quiz/Assessment Answers

Question 1: You are the manager of a startup energy provider. You have a variety of unstructured and structured data from your customers that you want to organize, including correspondence emails, customer zip codes, phone numbers, average energy consumption information, and copies of letters sent to customers. How can unstructured data be defined?

  • Data that cannot be easily stored or queried in a relational database
  • Data that only exists in a text format and is not numerical
  • All data that exists in a data lake and has not yet been processed
  • Data that can be recorded in a field that exists in a tabular format database

Question 2: You manage a website that provides users with personalized fashion and style advice. You do this through a machine learning model that is given a user’s style preferences, and the algorithm recommends clothing available from various websites. You receive a percentage of the sale if a customer decides to purchase the items. What are the benefits of personalization in machine learning in this use case?

  • It allows the system to provide more valuable services and scalability.
  • It allows the business to make more money when users utilize the service.
  • It allows the customer to engage in the service with full confidence in its ability.
  • It allows the system to recommend a product that it knows customers already own.

Question 3: Machine learning is becoming more affordable for businesses of all sizes. However, it can be tempting to use only one machine learning model in your business to save on costs, time, and maintenance. Why is it a good idea to divide a large machine learning use case into smaller use cases and then build separate models for each case?

  • Combining multiple small models is more powerful than using one model to do everything.
  • Multiple small models are less likely to encounter issues and faults than one model is.
  • Gathering training data for multiple small models is easier than gathering data for one larger model.
  • Using multiple small models saves on cost in comparison to one all-encompassing model.

Question 4: You work in a large technology company that has decided to create an app which uses the cloud to provide users with photo and image storage. You want to use machine learning to add features to the app such as search capabilities, automatic filters, facial recognition, subject recognition, and automatic album creation. To create such an app with these features would require many machine learning models working together. Which machine learning models would you expect to be used to automatically curate an album named “Cooking From Home”?

  • Location detection and subject image recognition
  • Image filters and blurry image detection
  • Text recognition and auto enhance features
  • Duplicate photo detection and food item classification

Question 5: You run a furniture store that accepts and sells used furniture. Until now, you have asked traders and employees to inspect the items and judge the condition of various parts. This information would be used to create a cost evaluation for the item. However, this business process is time consuming and laborious and is prone to oversights. How can you best use machine learning to automate the business processes in this scenario?

  • Train a machine learning model that can predict the cost of furniture based on specific photos.
  • Train a machine learning model that can predict the cost of furniture based on its color and size.
  • Train a machine learning model that can predict the cost of furniture based on traders’ and employees’ specific inspection information.
  • Train a machine learning model that tells traders and employees which item features should be inspected for more accurate evaluations.

Managing ML projects successfully Coursera Quiz/Assessment Answers

Question 1: You work in a company that is using legacy systems for data storage. Your current business processes occur slowly and rely on sharing data. You want to implement data lake modernization and make better use of data warehouses, but first you need to convince your teams and machine learning engineers to support this change. What current problems might teams and machine learning engineers experience while using traditional data storage solutions?

  • Inseparable combination of compute and storage
  • Poor performance
  • Having a centralized metadata store
  • No siloed access policies or technology

Question 2: You run a machine learning project at a startup gaming app development business. You want to improve the user experience of your games and increase the sales of microtransactions. The business has some data that they can use to train a machine learning model but not enough to create the accurate results you want. How can you enrich and improve your existing dataset before training your machine learning model?

  • Join your existing data with third-party data.
  • Use 100% of your dataset for training.
  • Add additional, but less relevant, existing data to your dataset.
  • Duplicate your dataset, which allows more data for training and evaluation.

Question 3: You work at an international sales organization and are planning a machine learning project. You are in the process of gathering and organizing customer data but have discovered that it is traditionally stored in offices around the world. You have contacted the global offices, which are willing to share their data with you but have no method of doing so. What is the best solution you could suggest that will help gather data for your project and also for other projects and processes in the future?

  • Convert data to digital formats while it still exists in data silos.
  • Move all the company’s storage data to one office in the world.
  • Move all the customer data into one legacy storage system at one location.
  • Convert data to digital formats and place customer data into one database.

Question 4: You manage a large e-commerce store with various departments that use sensitive customer data to improve business efficiency and processes. A customer can store their payment card details for faster online checkout. Also, the last 4 digits of a payment card are used to confirm a customer’s identity if they need to contact the customer service team. What technique can be used to protect payment card details when the customer service team needs to confirm a customer’s identity?

  • Redact all but the last 4 digits of a customer’s payment card details from the system.
  • Delete all but the last 4 digits of a customer’s payment card details from the system.
  • Change each digit of the customer’s payment card details to 0 or 5, whichever is closest.
  • Encrypt the customer’s full payment card details, which can only be decrypted with the last 4 digits of a payment card.

Question 5: You have convinced the decision makers at your small business to allow a pilot machine learning project. The leaders have provided a reasonable, but not high, budget for the project. Also, they do not want to spend too much of the employees’ time, which might affect the core business. The team you have formed is excited about the project but currently lacks the skills needed for the pilot. How can you best garner the expertise needed for the machine learning project?

  • Up-skill your current team using online learning platforms.
  • Borrow the sole IT analyst from another department in the business.
  • Hire new employees who already have the skills needed for this project.
  • Ask your team to self-research and study the skills needed for the project.

Leave a Comment