SYLLABUS PEC-CSE- 320-G DATA SCIENCE Class Work : 25 Marks Exam +75 Marks Total : 100 Marks Duration of Exam. : 3 Hrs, UNIT-I Introduction to Data Science: Concept of Data Science, Analysis vs Reporting, Collection, storing, modelling ar , Traits of Big data, Web Scraping, processing, describing and modelling, statistical ind algorithm modelling, A and data science, Myths of Data science. UNIT-I Introduction to Programming Tools for Data Science: Toolkits using Python: Matplotlib, NumPy, Scikit-learn, NLTK, Visualizing Data: Bar Charts, Line Charts, Scatterplots, Working with data: Reading Files, Scraping the Web. UNIT-IIL Data Science Methodology: Business Understanding, Analytic Approach, Data Requirements, Data Collection, Data Understanding, data Preparation, Model Deployment, feedback. UNIT-IV Data Science Application: Prediction and elections, Recommendations and business analytics, clustering and text analytics. NOTE : Examiner will set nine questions in total, Question one We be compulsory. Question one will have 6 parts of 2.5 marks ench from all bas: Temaining eight questions of 15 marks each (o be set by taking twoquestions rom cach unit. The students have to attempt five questions in total, first being compulsory and selecting one from cach unit. © scanned with OKEN ScannerJuly - 2021 Paper Code:-PEC-CSE-320-G ——— . Note: Attempt, five questions in all, selecting one question from each Section. Question No. 1 is compulsory. All questions carry equal marks. Q.1.(a) What are the differences between supervised and unsupervised Learning? Ans. Aspect [Supervised Learning Unsupervised Learning Learning typelLeams from labeled data__| Learns from unlabeled data Inputdata __|Input-output pairs Input data without labels Objective [Predict output based on input | Discover pattems or structure in data Feedback Error between predicted and_ | None true outputs Examples [Email spam detection, Image] Clustering, Dimensionality reduction classification Q.1.(b) What is AI? What is use of Al in data science? ‘Ans. Artificial intelligence (AI) isa rapidly evolving field that uses algorithms to enable robots to complete tasks automatically, replicating the cognitive functions associated with human and animal intelligence. Al models are created to recognize trends in previous data allowing for the automation of related tasks when similar pattems arise. ‘Al is widely used ina variety of industries, with huge organizations, including tech behemoths ike Facebook, Amazon, and Google, prinirily depending onto drive innovation, improve efficiency, and unleash new possibilities. Antificial intelligence is important in the subject of datascience, contributing to subfields such as programming, mathematics, and statistics. The ability to recognize patterns and trends in data is required for a Data Scientist to flourish in their role. The Data Science pipeline is defined by many important procedures. Q.1.(©) Explain the Myth of Data Sciences, i. « ‘Ans. Myth 1: Misconception about Mathematical Expertis widespread misconception that data science is only for mathematicians, While understanding statistics and probability is essential, the misperception stems from the idea that data scientists use sophisticated mathematical equations frequently in their day-to-day work. Myth 2: Tool Proficiency Equals Data Science Mastery : A prevalent fallacy is © scanned with OKEN Scannera . D, ata Scien. thar masteringa too) suchas SAS, correlates tohaving daa science knowledge, Whi atoolis beneficial it does not automatically qualify someone as adata scientist, "thy Myth 3: Impending Replacement of Data Scientists by AI: There is perception that artificial intelligence will completely replace data scientists, Hove S°! science improves automation may handle certain manual jobs, but humaninterege 2" by the appropriate information, remains essential fordireetingand contextaspea operations, alizing ALary, Myth 4: Perceived Difficulty in Adopting Data Seience : misconception that incomporating data science ino organizational procedures and problematic. Some businesses are resistant to adoption beeen we difficult. y Q.1.(d) Discuss Bar Charts, ‘Ans. Bar Plot : Bar Chart Use to show the Categorical data in bars where the heights of bars illustrate the value. In Matplotlib, a bar graph shows categorical data in the form of rectangular bars the lengths of which are proportional to their values. It is generated with the help of the bar() function; each bar stands for a category; the height ofa bar is undefined by the Tespective numbers Q.1.(e) Discuss the various data science applications. Ans. 1. Image and Speech Recognition — Data science enables machines to interpret images and understand spoken language through deep learning techniques. " 2. Gaming Industry — It enhances user experience and game design by analyzing player behavior and optimizing game mechanics. 3. Internet Search — Search engines use data science to rank results based on relevance, user behavior, and content analysis. pone ee 4, Transportation — Data science helps in route optimization, traffic prediction, anc 2fficient fleet management. See eee i and 5. Healthcare — It supports disease prediction, medical imaging analysis, a! sersonalized treatment recommendations. a © scanned with OKEN Scanner[Link] 6% Semester, Solved papers July -2021 2 Q.2.(a) Define Data science? List the differen ' unsupervised learning. _ Ans. Data Science ie field that combines skills from statistics, computer science, and machine learning to analyze both organized (like tables) and unorganized data (like images or text), Its goal is to find patterns, make predictions, and solve problems meni i Supervised Learning: In supervised learning, the algorithm leams from labeled data where cach example in the training dataset is associated with a corresponding label or output The algorithm aims to learn a mapping from inputs to outputs, making predictions or decisions based on the input features. Through repeated exposure to labeled data, the algorithm adjusts its parameters to minimize the difference between its predictions and the true outputs. In unsupervised learning, the algorithm learns from unlabeled data, where no explicit output labels are provided. Instead, the algorithm explores data to identify patterns, structures, orrelationships among the input features. Unsupervised learning algorithms aim to discover hidden patterns or groupings within the data without guidance or supervision. Clustering and dimensionality reduction are common tasks in unsupervised leaming. ces between supervised and ‘Aspect Supervised Learning Unsupervised Learning Learning typ¢ Learns from labeled data Leams from unlabeled data Input data —_| Input-output pairs Input data without labels Objective | Predict output based on input | Discover pattems or structure in datd Feedback | Errorbetween predicted and | None true outputs Examples | Email spam detection, Image | Clustering, Dimensionality reduction classification Training data] Requires labeled data for | Does not require labeled data for training training Evaluation | Accuracy, Precision, Recall Cluster quality, Dimensionality reduction Application | Classification, Regression _ | Clustering, Anomaly detection, Association Q.2(b) Define big data and explain and discuss the characteristics of big data and application of big data. ‘Ans, “Big data” refers to “high-volume, high-velocity, and complex information assets that demand cost-effective, innovative forms of information processing to optimize insight and decision making,” Big datas typically large volume of unstructured or semi structures and structured data that gets created from various organized and unorganized. application activities and channels such as Emails, Twitter, web blogs, Face book etc. © scanned with OKEN ScannerTe Data Sciny,, Example: Big data is more than TB per day For instance, consider the massive world of Facebook data in the context of Big Dany insocial media. Facebook isa perfect example ofthe massive data quantities generated dai, the social media world, According to Facebook's statistics, a whopping 500 terabytes, (TB) of new data enters the company’s database every day Characteristics of Big Data are: 1, Variety: Big Data Variety contains structured, unstructured, and semi-structured data obtained from various sources. 2. Velocity: Velocity is defined as the rate at which data is generated in real time. encompasses the velocity of change, the connecting of incoming data sets at variable speeds, and bursts « “activity in general. 3. Volume: One of the primary properties of big data is volume. Big Data, as the name suggests, refers to huge ‘volumes’ of data generated daily from many sources such as social media platforms, business processes, machines, networks, and human interactions. 4. Veracity: The reliability and accuracy of the data are referred to as its veracity. In the world of big data, where information is sourced from various channels and in various formats, assuring data integrity and trustworthiness is critical. 5. Value: While not one of the original ‘V's, the concept of Value is critical to understanding the significance of big data. Applications of Big Data: 1. Healthcare Analytics — Analyzing patient records and treatment outcomes to improve care and reduce costs. 2. Fraud Detection in Banking Real-time monitoring of transactions to detect unusual pattems and prevent fraud. 3. Smart Cities - Managing traffic, utilities, and emergency services using data from sensors and IoT devices. . 4. Retail Personalization — Delivering targeted promotions and improving customer experience based on shopping habits. . 5. Predictive Maintenance in Manufacturing — Monitoring equipment data to predict failures and schedule timely maintenance. Q.3(a) What is data preprocessing? Explain and discuss in detail the various steps involved in the data preprocessing. Ans, Preprocessing of data includes normalizing the data, encoding the data according to category, scaling the feature and data splitting, This step is important when trying to get the best accuracy from a developed model. - Feature Engineering: Transforming of the already existing data into new features that could potentially result in better performing models, = ik Data Integration: The process of linking various databases so as to come up witha Single integrated database. © scanned with OKEN Scanner= Blech 6" Semester, Solved papers, July -2 ly -2021 5 . - Data Reduction make computations faster and improve the results of a model, again data dimensionality needs be reduced using methods like the PCA. : Importane : -Data Quality p Increases the probability of obtaining accurate and reliable results of analysis and modeling by using competent quality data, _ = Model Performance: Helps in improving model accuracy sit frees data from many ofthe imperfections that may be present in the real world. 7 - Efficiency: Getting data quality issues out early reduces the number of cycles that ngeds tobe made through the analysis phase thus saving much time and ess expensive resources Inconclusion, data collection and exploration are major and essential steps of data science process. Data source and types, data collection techniques, exploratory data analysis and data cleaning and preprocessing help a data scientist to use quality data for constructing his/her analysis and arrive at a better insight of the truth. Q.3(b) What is stastical modeling and how is it used? what are the reasons to earn statistical modeling? Discuss important statistical techniques in data analysis. ‘Ans. Statistical modeling is a mathematical framework: used to represent relationships ‘among variables. It involves using: statistical methods to analyze data, make predictions, or infer patterns from the data. In simple terms, statistical modeling is about developing ‘model that fan describe the underlying structure of the data and make informed decisions or predictions based on it. These models are used to estimate relationships, test hypotheses, and predict future outcomes. Reasons (o learn statistical modeling : It helps to represent relationships between variables, test hypotheses, and forecast future events. It helps businesses and organizations make data-driven decisions, improve predictions, understand complex relationships, and solve problems effectively. Learning: statistical modeling is important because it improves decision- making accuracy, helps in understanding data, and enables more reliable predictions. Important Statistical Techniques in Data Analysis: 1. Regression Analysis: Helps predict outcomes by analyzing relationships between variables (¢.g., predicting sales based on advertisingspend). : / 2. Hypothesis Testing: Tests assumptions (e.g., comparing Wo marketing strategies tosee which is better). . ; , 3. Time Series Analysis: Analyzes data over time (¢.8-5 forecasting stock prices or Weather). : i 4, Bayesian Analysis: Updates predictions as new data comes in (@-8 adjusting, Predictions based on new information). & Clustering: Groups similar data together (¢¢., customer segmentation). 6 Principal Component Analysis (PCA): Re luces data complexity by focusing on key factors (e.g., simplifying survey data). © scanned with OKEN Scanner7) Data Stir, 7, Survival Analysis: Studies the time until an event happens (.g., patient 5 Uv 8. Correlation: Measures relationships between variables (e.g., income and. ed level). 7 9. Factor Analysi preferences). Finds hidden factors in data (e.g, understanding custon ___ Q4la)What is Python Matplotlib? What is Matplotlib used for? What are the ith eee ae of the chart? Discuss the various types of plots with suitable so Cs Marplotlib, also known as Mac OS X matplotlib, is an open-source Python 2D i ea a ded thatis used in many scientific applications such as data analysis and ‘computational Key components of Matplotlib include (1) Figure: A Figure serves as a canvas that can accommodate one or multiple axes (plots). It provides a space for organizing and displaying plots, (2) Axes: An individual plot within a Figure is referred to as an Axes. A Figure can contain multiple Axes, each with its own title, x-label, and y-label. Axes are the primary components where data visualization occurs. @)Axis: Axes consist of Axis objects responsible for generating graph limits and tick marks along the x and y axes. They determine the scale and range of the plotted data, (4) Artist: Artists encompass all graphical elements visible on the plot, such as Text objects, Line2D objects, and collections. Most Artists are linked with Axes and play an important role for displaying the Graphical illustration of data. Matplotlib is a powerful tool for data visualization in Python and plays acrucial role with application in the process of data exploration, analysis, and report of the results. Types of Charts : Data visualization encompasses various types of graphical representations that aid in understanding and interpreting data. (1) Bar Plot : Bar Chart Use to show the Categorical data in bars where the heights of bars illustrate the value. In Matplotlib, a bar graph shows categorical data in the form of rectangular bars the lengths of which are proportional to their values. (2) Line Charts : Line Charts A line plotting in Matplotlib isa mere graphing ofthe simplest nature where numerical values on a vertical axis or * Y-axis’ are plotted in relation to its horizontal axis or ‘X-axis’, as selected data points that are represented by these values are joined with lines, As (3) Pie Chart : The pie chart of the matplotlib is display the data distribution in the form of slices ofa circular plate. Originally the slices corresponding to each category and their sizes are depicted proportional to the size of the entire set of data. © scanned with OKEN Scanner[Link] 6" Semester, Solved papers, July -2021 7 Pie Chat Example category a category 0 category 8 __ @) Histogram : The Histogram plots drawn using Mapai in which the mamerical datais quantised into bins and the frequency of observations in each bin is determined. ogram Example (5) Heatmaps : In Matplotlib, a heatmap is defined as one type of graphical interpretation of data in which values that exist in a matrix are depicted as colors. It is often used to describe the relations or the distributions ‘of data into two dimensions. Scatter Plot Example Yass ° ans © scanned with OKEN Scanner8 Data Science (6) Seatter Plot : Matplotlib which isa common plots and graphs platform specifically uses the scatter plot to display the relationship of wo numerical variables where every single point is plotted on the xy-line. Q.4.(b) What isan array and how is it different from a list? What is the name of the built-in array class in NumPy? Create the following NumPy arrays. (i) a ID array called zeros having 10 elements and all the elements are set to zero (ii) a 1D array called vowels having the element ‘a’, ‘e’, 4’, ‘0’, ‘u’. (ii) A 2-D array called ones having 2 rows and 5 columns and all the element are set to 1 and dtype as int. Ans. Itisa Python library that takes the name Numerical Python abbreviated as NumPy. NumPy isan essential tool that is used in the language and which is created for carrying out arithmetic and numeric computations. Feature Python List NumPy Array 1 Library Part of the Python standard | Requires NumPy library library. 2 Creation Created using square Created using ‘[Link]()* brackets ‘[]‘. function or other array-generating| functions 3 DataType | Canstore elements of Usually stores elements of the sam Uniformity | different data types data type (homogeneous) 4 | Performance | Slower, especially for Faster and more efficient, large datasets. especially for large datasets due t vectorized operations 5 | Indexingand | Supports indexingand slicing} Supports indexing, slicing, and Slicing multidimensional slicing 6 | Example “python my_list= “python import numpy as np 0, 2,3, 4, 5)“ my_array = [Link]({1, 2, 3, 4, sy" For example, import numpy as np (i)a 1D array called ‘zeros’ having 10 elements, all set to zero zeros = [Link](10) (ii)a 1D array called ‘vowels’ having the elements ‘a’, ‘e’, “i”, ‘0°, ‘w’ vowels = [Link]([‘a’, ‘e’, ‘i’, ‘0’, ‘u’)) (iii) a 2D array called ‘ones’ having 2 rows and 5 columns, all set to 1 and dtype as int ee - © scanned with OKEN Scannerplech 6" Semester, Solved papers, July -202 7 ~ 9 e=int) ray *Zer0s":", zeros) print(“Array ‘vowels’:", vowels) print(“Array ‘ones’:", ones) Output Array “zeros’: (0.0. [Link].[Link].] Array ‘vowels’: [‘a" *e” io" “u] Array ‘ones’: (11111) gaia Q.5(a) Plot the following data on line chart: Day Income Monday 510 Tuesday 350 Wednesday 475 Thursday 580 Friday 600 (a) Write the title of the chart “The Weekly Income Report” (b) Write the appropriate titles of both the axes (c) Write code to Display legends (d) Display red color of the line (e) Use the line styled-dashed : (f) Display diamond styl markers on data points Ans. import [Link] plt days = [“Monday’ “Thursday”, “Friday”] income = [510, 350, 475, 580, 600] # Plotting the line chart [Link](days, income, color="red’, linestyle="— Income’) # Adding titles and labels [Link](“The Weekly Income Report’) [Link](‘Day") plt,ylabel(‘Income’) # Displaying legends pltlegendO) # Display the plot pltshow0 sMarker="D" label" Weekly © scanned with OKEN Scanner0 1 REESE Data Science The Weekly Income Report 600 | ~@- Weekly Income 350 500 9 E 2 = 450 350 Monday Tuesday © Wednesday Thursday Friday Dey Q.5(b) What do you means by file? list out the basic file modes available. write a statement to create a [Link] file with the following text. (i python file handling is very interesting and useful (ii) This is a text file created through python Ans. A file isa collection of data or information that is stored on a computer or other storage devices. Files can contain text, images, videos, or any other form of data. Files are used to store information in a persistent way, which means the data remains saved even after the computer is turned off. The basic file modes available in Python are: 1. ‘r’ (Read mode (default)): Opens the file for reading. If the file does not exist, it throws an error. 2. Sw? (Write mode): Opens the file for writing. If the file exists, it is overwritten. Ifthe file does not exist, a new one is created. 3. ‘a’—(Append mode); Opens the file for appending, Data is written at the end of the file. Ifthe file does not exist, it is created. 4, - (Exclusive creation): Creates a new file, but only if the file does not already exist. If the file exists, an error is thrown. 5. ‘b’ — (Binary mode): Opens the file in binary mode (for non-text files, like images). 6. *t’—(Text mode (default): Opens the file in text mode (for text files). © scanned with OKEN Scanner[Link] 6" Semester, Solved papers, July -22] # Open the file in write mode Cw’) with open(‘[Link]’, Ww’) as fi file. [Link] eC‘python file handling is very interesting and useful\n’ ) ("This is ate file ereated through python\n’y Q.6(a) What is business understandin understandi portant in data science? process in deta ig in data science? Why business iscuss business understand phases and } Ans. Business understanding in data science refers to the process of clearly defining _ the objectives, goals, and constraints of a data science project from a business perspective. It ensures that the technical data work is aligned with organizational priorities and delivers real, tionable value. Why Business Understanding is Important in Data Science? - Goal Alignment: Ensures data science efforts directly support business goals. - Problem Clarity: Helps identify what problem needs to be solved and how success jill be measured. ~ Resource Optimization: Saves time and effort by focusing on relevant data and Is. - Communication Bridge: Connects technical data teams with non-technical holders. Phases and Process of Business Understanding : Business understanding is typically the first phase of the Cross-Industry Standard Process for Data Mining (CRISP-DM). It includes the following sub-phases: 1. Determine Business Objectives: Identify the business problem or opportunity. 2. Assess the Situation: Evaluate the current business scenario, including available resources, data, risks, and assumptions. 3. Define Data Science Goals: Translate the business objectives into specific data science tasks. 4, Produce a Project Plan: Develop aclear plan detailing, Business understanding is the foundation of every successful data science project. Ithelps ensure the right problems are solved using the right data, in ways that deliver real value to the organization. Without strong business understanding, even the most advanced models can miss the mark, . Q.6(b) How to prepare data is used from Modeling to Evaluation. ‘Ans, Data preparation isa critical step in the data science workflow, especially when transitioning from modeling to evaluation. Properly prepared data ensures that the models built ateaccurate, reliable, and generalizable. Here's how data is prepared during this phase: 1. Data Cleaning: Remove duplicates, handle missing values, and correct inconsistent data, Ensures that the data used for modeling is accurate and consistent. © scanned with OKEN Scanner2 ‘ 2. Data Transformation: Makes the dat Data Science 3. Feature En; model performance. Sitti ataset: Divi into training. vaticees morse tine bed set: Dit ethe data into taining, validation and lest sets(e.g,, Training set is used to train the model, validation , 5. Data Balancing: Ensues fair model uainy jl final evaluation, 6. Model Traini 7. Hyperparameter Tuning: model parameters. 8. Model Evaluation: Test the model on unseen test data, 9. Model Monitoring and Updating: Co deployment to detect performance drops (model drift), From modeling to evaluation, data is cleaned, transformed, split, and fine-tuned to build a robust and adaptable model. Regular monitoring and updating ensure the model stays effective as data and conditions evolve. Compatible with machine leami ‘orithm: hin a neering: Create new features or modi ly e: ones mpra : ‘xisting (0 improve tinuously evaluate the model after Q.7(a) What is data requirement? What are the five methods of collecting data? How to prepare data from understanding to preparation. Ans. Data requirement refers to the specific types of data needed to solve a business Problem or support a decision-making process. It defines the quantity, quality, format, and source of the data that will be used in the data science project. Understanding data: requirements ensures that the right datas collected, analyzed, and interpreted to achieve the desired outcomes, Five Methods of Collecting Data : There are various ways to collect data, depending on the nature of the project. Here are five common methods: 1. Surveys and Questionnaires :Collect data directly from individuals through structured questions (e.g., online surveys, paper forms). 2. Interviews : Gather qualitative data through direct interaction with participants (e.g., one-on-one or group interviews), 3. Observations : Collect data by observing behaviors or phenomena in natural settings. 4. Existing Data Sources (Secondary Data): Utilize pre-existing datasets available from various sources (e.g., government databases, company records, social media). 5. Experiments and A/B Testing : Perform controlled experiments to collect data through randomized trials or A/B tests. How to Prepare Data from Understanding to Preparation : 1, Data Understanding: Gain a deep understanding of the available data, including its source, quality, and relevance to the problem. 2. Data Collection: Gather all the required data based on the defined data requirements and business objectives, 3. Data Cleaning: Improve data quality by identifying and rectifying errors, inconsistencies, and missing values, . © scanned with OKEN Scanner[Link] 6" Semester, Solved pape uly -202] 4. Data Transformation: Transform data i 5. Data Splitting: Split data into trainin; 13 intoa format suitable foranalysis and modeling, \g, validation, and test sets, : Q.7(b) Discuss the following in detail; (i) Analytic Approach Gi) Deployment and feedback Ans.(i) Analytic Approach : An Analytic Approach refers toa structured method used 10 solve problems or gain insights from data by applying logical reasoning, statistical techniques, and data science tools. Itis commonly used in fields such as business intellcence data science, healthcare, finance, marketing, and many more. om Steps in an Analytic Approach: 1, Problem Definition: Clearly define the problem you want to solve or the question you want to answer, 2, Data Collection: Gather relevant data from different sources like databases, APIs, surveys, oF sensors. 3. Data Cleaning and Preparation: Handle missing values, remove duplicates, fix errors, and convert data into a usable format. 4. Exploratory Data Analysis (EDA): Understand data patterns, correlations, and trends using descriptive statistics and visualizations. = 5. Model Selection and Development: Choose the right statistical or machine leaming model based on the probler 6. Model Evaluation: Test model accuracy using techniques like cross-validation, confusion matrix, or ROC-AUC curves. 7. Insights and Interpretati conclusions and actionable insights. 8. Decision-Making and Action:Use the insights to make informed decisions or act (eg, targeted marketing, cost-cutting, operational changes). 9. Monitoring and Feedback: Continuously track the results ofimplemented decisions and update the model as new data comes in. Interpret the model’s output to draw meaningful Ans.(ii) Deployment and feedback : Deployment is the stage where the validated Model is integrated into a production environment to deliver real-time predictions or insights. This could be through APIs, dashboards, mobile apps, or embedded systems. A well-structured deployment must consider: ; ~ Real-time or batch data processing, ~ System scalability, - Infrastructure robustness (cloud/on-premises), ~And continuous integration workflows. / Once deployed, real-time monitoring becomes crucial, Operational metrics such as Prediction accuracy, latency, throughput, and system load are tracked to ensure smooth © scanned with OKEN Scanner