19 Basic Machine Learning Interview Questions and Answers

There are several companies who hire data engineers or data scientists to make their data more reliable and secure; and for that purpose they use machine learning.

The companies may hire number of engineers who are data analyst, machine learning engineers, deep learning engineer.

All these posts are of similar job nature. The employer can ask different types of interview questions to hire the best employee for the company.

How can we solve the real world problems using machine learning? So we get some seniors and give the proper judgments.

Machine Learning Interview Questions and Answers

 1 – What is Machine learning?

Machine learning is the application of artificial intelligence which is programmed in such a way to access data and learn automatically to improve its experience.

The primary object of machine learning is to access/retrieve data and learn without the intervention of the human to make decisions.

2 – How will you teach machine learning in easy words?

The interviewer is interested that how you will explain the machine learning in easy words. How you describe the basic components with the help of examples.

There is an easy way to explain the machine learning with an example. When yours friend invites you in a party. You don’t know the participants in that party. You just classify all the participants after visualizing in gender, their age and dressing.

You have no prior knowledge or past knowledge and experience about participants in party which is known as un-supervised learning.

On the other side when you have knowledge about those participants you classify them in different groups is known as supervised learning.

3 – How many types of machine learning?

  There are three major types

  1. Supervised learning
  2. Un-Supervised learning
  3. Re-enforcement learning

4 – What is supervised learning?

The trained data is given to the machine to learn which is based on the characteristics and data sets. It is labeled data having groups on the basis of characteristics.

For example the shape and color of different fruits is given to the machine as training data. The machine will proceed and work in future on the basis of that given data.

5 – What is Un-Supervised learning?

When there is no data and information which is given to the computer. It is the learning without the teacher.

It collects and categorizes all the data on the basis of assumption and in groups. It groups the data on the basis of relationships and characteristics.

6 – What is Reinforcement learning?

This learning is based on environment or model. When you perform some action on machine it uses special software which leads to perform certain tasks.

The software has a specific model and having the steps of action to perform.   Example of reinforcement learning is playing game when an agent has a set of goals to get high score and feedback on basis of punishment and reward.

7 – What is deep learning?

Deep learning is not completely different from machine learning. Deep learning is the small part of machine learning.

Deep learning is based on neural networks. These neural networks are based on the idea of human brain. The inspiration came from the structure of human brain. It detects the features and working same as the human mind works.

8 – What is the neural networking?

Artificial neural network is an algorithm which allows computer or machine to learn by incorporating new data.

It works like human brain. Neuron is the main object of human brain. It works like the same way.

9 – What is classification and regression in machine learning?

Both classification and regression are the part of supervised learning. When you predict continuous values like predicting stock market and try to predict sales.

Classification is based on the class to predict whether customers is going to buy some product or not and salary is predicted as high or low. It classifies in labels on the basis of characteristics.

10 – What do you understand by selection bias?

In statistical terms, bias is the sampling of data on the basis of population. Take an example, when you want to get information about the use of gaming computers in some specific state. To get accurate information you have to take data from all the prevailing markets that are dealing with gaming computers in that state.

If you assume to get data from one city you can be called bias on the collection of data. You are not collecting the data from all over the state. This may produce wrong conclusion.

11 – What is precision and recall?

Recall is the process of recall previous events which is held or managed by you. For example if your friend is giving you gifts on your birthday from last ten years.

One day your friend asks you to remember about the all gifts given on birthdays, then you recall all the previous birthdays events and try to remember about the gifts means recalling memory.

When you recall your memory, you may answer it right or wrong.  The precision is the ratio of a number of events you can correctly recall. If you recall 8 out of 10 birthday events then precision is 80%.

12 – What are true positive, true negative, false positive and false negative?

Let take an example to understand above terms. We have a model in which alarm goes on or not in case of fire or otherwise.

True positive:

If the alarm goes on in case of fire it is known as true positive. In this case, the fire is positive and prediction made by system to alarm is true.

False Positive:

If alarm goes on when there is no fire, in this situation fire is positive and the prediction made by the system is false. This is the worst condition.

True Negative:

If alarm does not go on when there is no fire. System considered the fire as negative and prediction made by the system is true.

False Negative:

If the alarm does not go on when there is fire. System considered fire as negative and prediction made by the system is false.

See also: 42 Data Science Interview Questions & Answers

13 – What is confusion Matrix?

A model have matrix which is used to make predictions. It is also known as error matrix which is designed in tables for easy identifications but its terminology looked confusing.

14 – What is inductive learning and deductive learning?

Inductive learning is the learning in which the learner discovers rules from specific to general phenomena. Based on some examples a learner can get into conclusion.

The deductive learning is the learning in which learners have some specific rules from conclusion and get specific observations. It works more general to more specific.

15 – What is clustering in machine learning?

The method of identifying similar groups of data in one data set is called clustering.

In other words it is the process of making different groups on the base of data structure.

Similar type of data is put in one group or cluster. For example a retailer wants to improve its business and try to gets reviews from different customers. All reviews are categorized in different possible groups called clusters to put suggestion to improve the business.

16 – What is KNN Clustering and K-means clustering?

KNN stands for K-Nearest Neighbor. It is used in supervised learning technique. This algorithm uses the method of classification or regression to make the clustering on continuous values.

K-means clustering is un-supervised learning technique. It is used in clustering. This is an algorithm to make classification on the basis of attributes of features.

17 -What ROC curve, how and when you use this also the representation?

ROC curve stands for Receiver Operating Characteristic curve. It is the fundamental tool to diagnose the testing of algorithm in machine learning.

It tests the algorithm to specify the true positive rate and false positive rate. The more area this curve takes, the better algorithm it is. The true positive rate should increase faster in this curve for algorithms.

18 – What is difference between in type-I and type-II error?

Type-I error is false positive. When algorithm specifies something which actually can’t be true and model shows that it is true. For example algorithm shows that a male person is pregnant. It is the example of false positive which will never happen.

Type-II error is false negative error in which the machine shows false results. For example if a woman is pregnant and the machine shows that the woman is not pregnant, then this algorithm has some error.

19 – What is more important; model accuracy or model performance?

Model accuracy is the part of model performance. It is sub set of model performance. For example if there are bulk of data and set of rows and system have to identify the fraud in this data.

It will happen through this model accuracy that should be higher to increase the model performance.

machine learning interview questions and answers
Machine learning interview questions and answers

3 Comments

Leave a Reply