What exactly is R?

 

R LANGUAGE

R

is a programming language that you may make as helpful as you wish. It's a tool you have at your disposal that can be used for many things, including    statistical                   analysis,    data           visualization,             data   manipulation, predictive modelling, forecast analysis, etc. Google, Facebook, and Twitter

are among the firms that use R.

Interview Questions on R

1.   What exactly is R?

R is a free and open-source programming language and environment for statistical computation and analysis or data science.

2.    What are the various data structures available in R? Explain them in a few words.

These are the data structures that are available in R:

Vector

A vector is a collection of data objects with the same fundamental type, and components are the members of a vector.

Lists

Lists are R objects that include items of various types, such as integers, texts, vectors, or another list.

Matrix

A matrix is a data structure with two dimensions, and vectors of the same length are bound together using matrices. A matrix's elements must all be of

the same type (numeric, logical, character).

DataFrame

A data frame, unlike a matrix, is more general in that individual columns might contain various data types (numeric, character, logical, etc.). It is a rectangular list that combines the properties of matrices and lists.

3.   What are some of the advantages of R?

Understanding the benefits and drawbacks of various languages and ecosystems is critical, and R is no different. So, what are the benefits of using R?

 

It is open-source. For different reasons, this counts as both a benefit and a defect, but being open source means it's publicly available, free to use, and expandable.

Its ecosystem of packages. As a data scientist, you don't have to spend a lot of time recreating the wheel, thanks to the built-in functions provided by R packages.

Its statistical and graphical abilities. R's graphing skills, according to many people, are unrivaled.

4.   What are the disadvantages of using R?

You should be aware of the drawbacks of R, just as you should know its benefits.

 

Memory and ability to perform. R is often compared to Python as the less powerful language in memory and performance. This is debatable, and many believe it is no longer relevant now that 64-bit systems have taken over the market.

It's free and open source. Open-source software offers both pros and cons. There is no governing organization in charge of R. Therefore, there is no single point of contact for assistance or quality assurance. This also implies that the R packages aren't always of the best quality.

Security. Because R was not designed with security in mind, it must rely on third-party resources to fill in the holes.

5.   How do you import a CSV file?

It's simple to load a.csv file into R. You have to call the "read.csv()" method and provide it with the file's location .house<- read.csv("C:/Users/John/Desktop/house.csv")

6.   What are the various components of graphic grammar?

There are, in general, several components of graphic grammar: Facet layer

Themes layer

Geometry layer Data layer

Co-ordinate layer Aesthetics layer

7.   What is Rmarkdown, and how does it work? What's the point of it?

RMarkdown is an R-provided reporting tool. Rmarkdown allows you to produce high-quality reports from your R code. Rmarkdown may produce the following output formats:

 

HTML PDF WORD

8.   What is the procedure for installing a package in R? To install a package in R, do the following command: install.packages(“<package name>”)

9.   Name a few R programs that can be used for data imputation?

These are some R packages that may be used to input data.

 

MICE

Amelia MissForest

Hmisc Mi imputeR

10.   Can you explain what a confusion matrix is in R?

A confusion matrix can be used to evaluate the model's accuracy. A cross- tabulation of observed and anticipated classes is calculated. The "confusionmatrix()" function from the "caTools" package can be used for this.

11.  List some of the functions in the "dplyr" package

The dplyr package includes the following functions: Filter

Select

Mutate Arrange

 

Count

12.   What would you do if you had to make a new R6 Class?

To begin, we'll need to develop an object template that contains the class's "Data Members" and "Class Functions."

These components make up an R6 object template: Private DataMembers

Name of the class

Functions of Public Members

13.   What do you know about the R package rattle?

Rattle is a popular R-based GUI for data mining. It provides statistical and visual summaries of data, converts data to be easily modeled, creates both unsupervised and supervised machine learning models from the data, visually displays model performance, and scores new datasets for production deployment. One of the most valuable features is that your

interactions with the graphical user interface are saved as an R script that can be run in R without using the Rattle interface.

14.   What are some R functions which can be used to debug?

The following functions can be used for debugging in R: traceback()

debug()

browser()

 

trace() recover()

15.   What exactly is a factor variable, and why would you use one?

A factor variable is a categorical variable that accepts numeric or character string values as input. The most important reason to employ a factor variable is that it may be used with great precision in statistical modeling. Another advantage is that they use less memory. To make a factor variable, use the factor() function.

16.   In R, what are the three different sorting algorithms?

R's sort() function is used to sort a vector or factor, mentioned and discussed below.

 

Radix: This non-comparative sorting method avoids overhead and is usually the most effective. It's a reliable algorithm used to calculate integer vectors and factors.

Quick Sort: According to R documentation, this function "uses Singleton (1969)'s implementation of Hoare's Quicksort technique and is only accessible when x is numeric (double or integer) and partial is NULL." It isn't regarded as a reliable method.

Shell: According to the R documentation, this approach "uses Shellsort (an O(n4/3) variation from Sedgewick (1986)."

17.   How can R help in data science?

R reduces time-consuming and graphically intense tasks to minutes and keystrokes. In reality, you're unlikely to come across R outside of the world of data science or a related discipline. It's useful for linear and nonlinear modeling, time-series analysis, graphing, grouping, and many other tasks.

Simply put, R was created to manipulate and visualize data. Thus it's only logical that it is used in data science.

18.   What is the purpose of the () function in R?

We use a () function to construct simpler code by applying an expression to a data set. Its syntax is as follows:

R Programming Syntax Basics

R is the most widely used language for statistical computing and data analysis, with over 10,000 free packages available in the CRAN library. Like any other programming language, R has a unique syntax that you must learn to utilize all of its robust features.

The R program's syntax

Variables, Comments, and Keywords are the three components of an R program. Variables are used to store data, Comments are used to make code more readable, and Keywords are reserved phrases that the compiler understands.

CSV files in R Programming

CSV files are text files in which each row's values are separated by a delimiter, such as a comma or a tab.

1.   Reading a CSV file

The read.csv(...) function in R may read the contents of a CSV file as a data frame. The CSV file must be read in the current working directory, or the directory must be established appropriately in R using the setwd(...) function. The read.csv() method may also read a CSV file via a URL.

2.   Using CSV files for querying

The R subset(csv data) function may extract the relevant result from SQL queries on the CSV content. Multiple queries can be run through the function simultaneously, separated by a logical operator. In R, the result is saved as a data frame.

3.   Inputting data into a CSV file

The data frame's contents can be saved as a CSV file. The CSV file is saved using the name supplied in R's write.csv(data frame, output CSV name) function in the current working directory.

Confusion Matrix

In R, a confusion matrix is a table that categorizes predictions about actual values. It has two dimensions, one of which will show the anticipated values, and the other will show the actual values.

Each row in the confusion matrix will represent the anticipated values, while the columns represent the actual values. This can also be the case in reverse. The nomenclature underlying the matrixes appears to be sophisticated, even though the matrices themselves are simple. There is always the possibility of being confused regarding the lessons. As a result, the phrase – Confusion matrix – was coined.

The 2×2 matrix in R was visible in the majority of the resources. It's worth noting, however, that you may make a matrix with any number of class values.

A confusion matrix is a value table representing the data points' predicted and actual values. You may use R packages like caret and gmodels and methods like a table() and crosstable() to understand your data better.

Interview Questions on Confusion Matrix

1.   What is the purpose of the confusion matrix? Which module do you think you'd use to demonstrate it?

A confusion matrix is one of the simplest ways to summarize the performance of your algorithm in machine learning. Judging the accuracy of a model by looking at the accuracy might be challenging at times

because of issues such as unequal distribution. Using a confusion matrix is a better approach to see how good your model is.

Let's start with some definitions.

Classification Accuracy: This is the ratio of the number of right predictions to the number of predictions.

True-positives: These are accurate predictions of real-world events.

False-positives: These are the predictions of actual events that turn incorrect.

True-negatives: True negatives are accurate forecasts of fake events.

False-negatives: False negatives are forecasts of false events that are incorrect.

The confusion matrix has been simplified to a list of true-positives, false- positives, true-negatives, and false-negatives.

2.   What is the definition of accuracy?

It's the most basic performance metric, and it's just the ratio of correctly predicted observations to total observations. We may say that it is best if our model is accurate. Yes, accuracy is a valuable statistic, but only when you have symmetric datasets with almost identical false positives and false negatives.

True-Positive + True-Negative / (True-Positive + False-Positive + False- Negative + True-Negative) / (True-Positive + False-Positive + False- Negative + True-Negative) / (True-Positive + False-Positive + False- Negative + True-Negative

3.   What is the definition of precision?

It's also referred to as the positive predictive value. In your predictive model, precision is the number of right positives as compared to the overall number of positives it forecasts.

True-Positives / (True-Positives + False-Positives) Precision = True- Positives / (True-Positives + False-Positives). True-Positives / Total Predicted Positives = Precision

It's the number of correctly predicted positive items divided by the total number of correctly predicted positive elements. Precision may be defined as a metric of exactness, quality, or correctness.

Exceptional accuracy: This indicates that most, if not all, of the good outcomes you predicted, are right.

4.   What is the definition of recall?

Recall that we may also refer to this as sensitivity or true-positive rate. The model predicts many positives compared to our data's actual number of positives.

True-Positives/(True-Positives + False-Positives) = Recall True-Positives / Total Actual Positives = Recall

A recall measures completeness. Our model had a high recall, implying it categorized most or all positive aspects as positive.

Random Forest in R

1.   What is your definition of Random Forest?

Random Forest is a form of ensemble learning approach for classification, regression, and other tasks related to Random Forests. Random Forests works by training a large number of decision trees simultaneously, and this is accomplished by averaging many decision trees from various portions of the same training set.

2.     What are the outputs of Random Forests for Classification and Regression problems?

 

Classification: The Random Forest's output is chosen by the most trees.

Regression: The mean or average forecast of the various trees is the Random Forest's output.

3.   What do Ensemble Methods entail?

Ensemble techniques are a machine learning methodology that integrates numerous base models to create a single best-fit prediction model. Random Forest are a form of ensemble method.

However, there is a law of decreasing returns in ensemble formation. The number of component classifiers in an ensemble significantly influences the accuracy of the prediction.

4.   What are some Random Forest hyperparameters?

Hyperparameters in Random Forest include:

 

The forest's total number of decision trees.

The number of characteristics that each tree considers while splitting a node.

The individual tree's maximum depth.

The minimum number of samples to divide at an internal node. The number of leaf nodes at its maximum.

The total number of random characteristics The bootstrapped dataset's size.

5.   How would you determine the Bootstrapped Dataset's ideal size?

Even though the size of the bootstrapped dataset is different, the datasets will be different since the observations are sampled with replacements.

As a result, the training data may be used in its entirety. The best thing to do most of the time is ignoring this hyperparameter.

6.   Is it necessary to prune Random Forest? Why do you think that is?

Pruning is a data compression method used in machine learning and search algorithms to minimize the size of decision trees by deleting non-critical and redundant elements of the tree.

Because it does not over-fit like a single decision tree, Random Forest typically does not require pruning. This occurs when the trees are

bootstrapped and numerous random trees employ random characteristics, resulting in robust individual trees not associated with one another.

7.   Is it required to use Random Forest with Cross-Validation?

A random forest's OOB is comparable to Cross-Validation, and as a result, cross-validation is not required. By default, random forest uses 2/3 of the data for training, the remainder for testing in regression, and about 70% for training and testing in classification. Because the variable selection is randomized during each tree split, it is not prone to overfitting like other models.

8.     What is the relationship between a Random Forest and Decision Trees?

Random forest is an ensemble learning approach that uses many decision trees to learn. A random forest may be used for classification and regression, and random forest outperforms decision trees and does not have the same tendency to overfit the data.

Overfitting occurs when a decision tree trained on a given dataset becomes too deep. Decision trees may be trained on multiple subsets of the training information to generate a random forest, and then the different decision trees can be averaged to reduce variation.

9.   Is Random Forest an Ensemble Algorithm?

Yes, Random Forest is a tree-based ensemble technique that relies on a set of random variables for each tree. Bagging is used as the ensemble approach, while decision tree is used as the individual model in Random Forest.

Random forests can be used for classification, regression, and other tasks in which a large number of decision trees are built at the same time. The random forest's output is the class most trees choose for classification tasks.

The mean or average forecast of the individual tresses is returned for regression tasks. Decision trees tend to overfit their training set, corrected by random forests.

K-MEANS Clustering

1.   What are some examples of k-Means Clustering applications?

The following are some examples of k-means clustering applications:

 

Document classification: Based on tags, subjects, and the document's substance, k-means may group documents into numerous groups.

 

Insurance fraud detection: It is feasible to identify new claims based on their closeness to clusters that signal fraudulent tendencies using previous data on fraudulent claims.

Criminals who use cyber-profiling: This is the practice of gathering data from people and groups to find significant correlations. Cyber profiling is based on criminal profiles, which offer information to the investigation division to categorize the sorts of criminals present at the crime scene.

2.       How can you tell the difference between KNN and K-means clustering?

The K-nearest neighbor algorithm is a supervised classification method known as KNN. This means categorizing an unlabeled data point, requiring labeled data. It tries to categorize a data point in the feature space based on its closeness to other K-data points.

K-means Clustering is a method for unsupervised classification. It merely needs a set of unlabeled points and a K-point threshold to collect and group data into K clusters.

3.   What is k-Means Clustering?

K-means Clustering is a vector quantization approach that divides a set of n observations into k clusters, with each observation belonging to the cluster with the closest mean. Within-cluster variances are minimized using k- means clustering.

Within-cluster-variance is an easy-to-understand compactness metric. Essentially, the goal is to split the data set into k divisions in the most

compact way possible.

4.   What is the Uniform Effect produced by k-Means Clustering?

The Uniform Effect refers to the tendency of k-means clustering to create clusters of uniform size. Even if the data behaves differently, uniform sizes ensure that the clusters have about the same number of observations.

5.   What are some k-Means Clustering Stopping Criteria?

The following are some of the most common reasons for stopping:

 

Convergence. There are no more modifications; the points remain in the same cluster.

The number of iterations that can be done. The method will be terminated after the maximum number of iterations has been reached. This is done to keep the algorithm's execution time to a minimum.

Variance hasn't increased by at least x%.

The variance did not increase by more than x times the starting variance.

MiniBatch k-means will not converge, so one of the other criteria is required. The number of iterations is the most common.

6.      Why does the Euclidean Distance metric dominate in k-Means Clustering?

The construction of k-means is not reliant on distances, and Within-cluster variance is decreased using K-means. When you examine the variance definition, you'll notice that it's the sum of squared Euclidean distances from the center.

The goal of k-means is to reduce squared errors. There is no such thing as "distance" in this case. Pairwise distances between data points are not explicitly used in the k-means process. It entails assigning points to the nearest centroid over and over again, based on the Euclidean distance between data points and a centroid.

Euclidean geometry is the origin of the term "centroid." In Euclidean space, it is a multivariate mean. Euclidean distances are the subject of Euclidean space. In most cases, non-Euclidean distances will not cross Euclidean space, which is why K-Means is only used for Euclidean distances.

Using arbitrary distances is incorrect because k-means may stop converging with other distance functions.