Interview Questions Set 1(51 to 60)

##datascience ##machinelearning ##statistics

Akash Borgalli Dec 30 2021 · 3 min read
Share this

51. AI Solution for architecture -- Let’s suppose there is an agricultural field in diff areas in India, and we know soil & weather condition is different over India, So I am trying to build a system that helps me understand what kind of treatments I will be able to apply on my crops, which crop I can grow in a particular month so I can be able to maximize the benefit from the soil. Then what kind of algorithm you will use whether it's ML, DL, Vision? What will be your approach and what kind of solution design you will provide?

Ans: I will first try to gather information from farmers considering climate, soil conditions, and market demand knowing what treatments they used to apply then based on that I'll try to prepare a dataset to train the machine learning model accordingly. I will also try to embed IoT solutions where from IoT sensors I will get climate and soil conditions in real-time after which I will try to build a multi-class classification ml model which will be trained on the dataset I have prepared and would predict which crops are suitable and what treatments need to be done on crops to the farmers located in specific regions of India notifying them by Email or SMS notification.

52. I have a client, they are facing a problem in terms of maintaining the pipeline for water. So what kind of AI solution you will design to identify the leakage and maintenance?

Ans: There would be several hydraulic models installed which could continuously send data and based on that data I will try to predict where the leakage can occur or the changes are of leakage after which the pipeline fixers will go to repair that pipeline in that area for accurate analysis we can also consider the type of pipes are in use and environment where pipes are installed into account while predicting leakage.

53. Let’s suppose I am building a solution for blind people What kind of AI solution you will provide to help them to interact with the system and an Affordable solution?

Ans: I will try to embed voice-enabled capabilities(speech services) so if they could not see they can interact with the AI system using their voice. Also, we can add AI System to convert text into speech or Braille in real-time for reading for blind persons. In Azure, we have something called an Immersive Reader which reads aloud the texts from a book or article you are reading. 

54. What is the difference between R2 and Adjusted R2?

Ans: The difference between R2 and Adjusted R2 is R2 gives results based on fixed-line whereas Adjusted R2 is a statistical approach where values are calculated mathematically.

The formula for R2:

Figure 1

The formula for Adjusted R2

Figure 2

55. Where do you apply Regularization and What kind of regularization you have applied and Why?

Ans: The regression model is a statistical procedure generally used to solve regression problems to avoid overfitting the model. In short, it is used to build a robust or stable model by estimating the linear or polynomial relationship between variables. I have used Elastic net to solve one of the linear regression problems as ElasticNet is the combination of L1(Lasso) and L2(Ridge) regularization techniques.

56. What do you understand by multicollinearity and homoscedasticity in Dataset?

Ans: When two features that are dependent variables from the dataset show some kind of relation or pattern then those features are said to be multicollinear. This is a concept of multicollinearity.

Homoscedasticity is said when there is a condition of having error term or variance of the residual in a regression model is constant. In the dataset, all the data points in a progression show the same finite variance.

57. Can you please explain 1 example of Polynomial Regression and how to build a model for polynomial regression.

Ans: It is used to solve a problem that does not shows linearity once plotted on a graph. In short, used to solve the non-collinear relationships between data. So before building the model I will try to transform the data into its polynomial form and then try to apply non-linear regression.

58. There is some client who is intercepting a call like 3, 4, or 5 people talking in zoom call. Tell me to approach so that we can able to separate the voices of each and every person. (Hint: Speech Diarization)

Ans: We will use speech recognization services which would help us identify the voices of the individual persons. So it will help us distinguish between the voices which would also give transcripts out of it as to which person said what.

59. In the case of the multilinear regression model, let’s suppose my number of features is 5. Can you explain to me what kind of line it draws? Explain.

Ans: It will draw 5 different best-fitted lines based on the particular individual features of a dataset.

60. List no. of algorithms that you know from clustering?

Ans: The algorithms that I know for clustering which is an unsupervised machine learning approach and those are K-Means, K-Means++, Hierarchical Mean, DBScan.

Read next