Computer Science · 10th Grade · Advanced Data Structures and Management · Weeks 10-18

Data Cleaning and Preprocessing

Students learn techniques for cleaning and preprocessing raw data to ensure its quality and suitability for analysis.

• Common Core State Standards
• CSTA: 3A-DA-10

About This Topic

Data cleaning and preprocessing is the unglamorous but critical foundation of any data analysis project. In US 10th-grade computer science, students encounter raw data that includes missing values, duplicate records, inconsistent formatting, and outliers that skew results. Learning to identify and resolve these issues prepares students for CSTA Standard 3A-DA-10, which emphasizes data collection, storage, and analysis with an eye toward quality and integrity.

When students work with genuinely messy datasets, they quickly discover that the same problem can be approached in multiple valid ways. Should a missing value be deleted, replaced with a mean, or flagged? These judgment calls require students to think about context and purpose, not just technical procedure.

Active learning is especially productive here because students can physically annotate printed datasets, argue over cleaning decisions in pairs, and present their rationale to the class. The social negotiation of 'what counts as clean' mirrors professional data team discussions and deepens conceptual understanding far beyond reading about it.

Key Questions

1. Explain the common types of data inconsistencies and errors.
2. Analyze the impact of dirty data on analytical results.
3. Construct a plan for cleaning a given messy dataset.

Learning Objectives

• Identify common data inconsistencies such as missing values, duplicate entries, and formatting errors in a given dataset.
• Analyze the impact of specific data quality issues, like outliers or incorrect data types, on statistical calculations and visualizations.
• Formulate a step-by-step plan to clean a provided messy dataset, justifying each cleaning decision.
• Evaluate the effectiveness of different data cleaning strategies for a specific analytical goal.
• Demonstrate the application of data cleaning techniques using a programming tool or spreadsheet software.

Before You Start

Introduction to Data Types

Why: Students need to understand basic data types (numerical, categorical, text) to identify data type mismatches.

Basic Spreadsheet or Programming Fundamentals

Why: Students require foundational skills in tools like Excel, Google Sheets, or Python libraries (like Pandas) to perform practical data cleaning operations.

Descriptive Statistics (Mean, Median, Mode)

Why: Understanding basic statistical measures is crucial for analyzing the impact of dirty data and for performing imputation.

Key Vocabulary

Missing Values	Data points that are absent or not recorded for a particular observation. These can be represented as blank cells, NA, or null.
Duplicate Records	Identical or near-identical entries for the same entity within a dataset. These can inflate counts and skew analysis.
Data Type Mismatch	Occurs when a column contains values that do not conform to the expected data type, such as text in a numerical field.
Outlier	A data point that significantly differs from other observations in the dataset. Outliers can be genuine extreme values or errors.
Data Imputation	The process of replacing missing data points with substituted values, such as the mean, median, or a predicted value.

Watch Out for These Misconceptions

Common Misconception

Cleaning data just means deleting rows with problems.

What to Teach Instead

Deletion is only one strategy and often the wrong one. Imputation, normalization, flagging, and transformation are equally valid approaches depending on context. When students defend their choices to peers in collaborative activities, they develop a richer toolkit and learn to match strategies to situations.

Common Misconception

Data errors are always obvious and easy to spot.

What to Teach Instead

Many data errors are subtle, such as a birth year entered as 1920 instead of 2012, or a city name spelled two different ways. Teaching students to use statistical summaries (min, max, unique counts) rather than visual scanning alone helps build the habit of systematic auditing.

Common Misconception

Preprocessing and analysis are separate phases that never overlap.

What to Teach Instead

In practice, analysts often discover new data quality issues during analysis and must return to preprocessing. Students benefit from understanding this iterative cycle rather than viewing cleaning as a one-time gate before the 'real' work begins.

Active Learning Ideas

Gallery Walk: The Messy Dataset Museum

Print five different messy datasets and post them around the room, each with a different type of data quality problem (duplicates, missing values, format mismatches, outliers, impossible values). Groups rotate through stations with sticky notes to identify the problem type and propose a cleaning strategy before moving on.

35 min·Small Groups

Think-Pair-Share: Should We Delete It?

Give students a dataset with 15% missing age values and ask them individually to decide whether to delete rows, fill with the mean, or flag the records. Pairs compare decisions and discuss trade-offs, then share cases where they disagreed and why.

20 min·Pairs

Inquiry Circle: Before-and-After Analysis

Small groups receive the same raw sales dataset and a pre-cleaned version. They must reverse-engineer which cleaning steps were applied by comparing the two versions, then write a short cleaning log documenting each transformation in order.

30 min·Small Groups

Structured Discussion: The Cost of Dirty Data

Share a real case study (e.g., a hospital billing error or a census miscoding) where uncleaned data led to a costly mistake. The class discusses what preprocessing step could have caught the error, then identifies which step from their cleaning toolkit would apply.

15 min·Whole Class

Real-World Connections

• Financial analysts at major banks meticulously clean transaction data to detect fraudulent activity, ensuring accurate reporting and preventing financial losses. Inaccurate data could lead to misidentification of suspicious patterns.
• Epidemiologists at the Centers for Disease Control and Prevention (CDC) clean patient data from various sources to track disease outbreaks. Inconsistent formatting or missing demographic information can hinder the timely identification of public health threats.
• Marketing teams at e-commerce companies clean customer databases to segment audiences for targeted advertising campaigns. Duplicate customer entries or incorrect contact information can lead to wasted marketing spend and customer frustration.

Assessment Ideas

Exit Ticket

Provide students with a small, messy dataset (e.g., a CSV snippet with errors). Ask them to identify two specific data quality issues present and suggest one cleaning step for each. Collect these as they leave class.

Quick Check

Present students with a scenario: 'A dataset of student test scores has missing scores for 10% of students and some scores are entered as text (e.g., 'ninety').' Ask them to list three potential problems this data could cause for calculating the class average and propose one method to address each problem.

Discussion Prompt

Pose the question: 'Imagine you are cleaning a dataset of product prices, and you find a price of $0.01 for a laptop and $1,000,000 for a pen. How would you decide if these are errors or valid extreme values? What factors would influence your decision?' Facilitate a class discussion on critical thinking in data cleaning.

Frequently Asked Questions

What are the most common types of data quality problems students encounter?

The most frequent issues are missing values, duplicate records, inconsistent formatting (e.g., 'USA' vs 'United States' vs 'US'), outliers that fall outside realistic ranges, and data type mismatches where numbers are stored as text. Students typically encounter all five types in a single real-world dataset.

How does dirty data affect the results of a data analysis?

Dirty data can distort every statistical measure. A single extreme outlier can pull a mean far from the true center. Duplicate records inflate counts. Missing values reduce sample size and can introduce bias if absences are not random. In machine learning models, dirty training data produces unreliable predictions regardless of model sophistication.

What does a data cleaning plan typically include?

A cleaning plan documents the original data source, a summary of identified issues, the chosen remedy for each issue type, the order in which transformations are applied, and a validation step confirming the cleaned data meets quality criteria. Keeping this log allows others to reproduce or audit the process.

How does active learning help students understand data preprocessing?

When students physically annotate a messy dataset and argue over cleaning decisions with a partner, they confront ambiguity that a clean textbook example hides. Justifying choices out loud reveals gaps in reasoning. Group disagreements about whether to delete or impute a value often generate the most durable learning about trade-offs.