bytes

tutorials

data science

normalization in machine learning

Module - 2 Data preparation and EDA

Lesson - 222 Data Normalization

**Overview**

Data Normalization is an vital pre-processing step in Machine Learning (ML) that makes a difference to make sure that all input parameters are scaled to a common range. It is a procedure that's utilized to progress the exactness and proficiency of ML algorithms by changing the information into a normal distribution.

**Introduction to Data Normalization:**

Data normalization in ML that transforms data into a common format, so it can be used in analytics and machine learning algorithms. It is typically used to transform the raw data into a more useful form for ML algorithms such as linear regression, logistic regression, and neural networks. Data Normalization can be applied to numerical and categorical data, and it can help to reduce the complexity of the data.

For illustration, it can offer assistance to diminish the number of features by combining diverse features into one or by expelling redundant features. It can moreover be utilized to standardize the information to guarantee that all input parameters are inside the same range. In expansion, it can offer assistance to decrease the impact of outliers and decrease the chances of overfitting.

**Types of Data:**

**Nominal data**is a type of data that is not ordered or ranked. It is usually qualitative in nature and can be used to categorize items. For example, hair color (blonde, brunette, black, etc.) is a type of nominal data. When normalizing nominal data, it's important to use techniques that don't rely on the numeric values of the data, such as one-hot encoding.**Ordinal data**is a type of data that is ordered or ranked. It is usually qualitative in nature and can be used to rank items. For example, a survey scale that asks respondents to rate something on a scale of 1 to 5 is an example of ordinal data. When normalizing ordinal data, it's important to use techniques that preserve the order of the data, such as min-max normalization.**Interval data**is a type of data that is ordered, but not necessarily ranked. It is usually quantitative in nature and can be used to measure the difference between two items. For example, temperature is an example of interval data. When normalizing interval data, it's important to use techniques that preserve the interval between the data points, such as z-score normalization.**Ratio data**is a type of data that is both ordered and ranked. It is usually quantitative in nature and can be used to measure the ratio between two items. For example, length is an example of ratio data. When normalizing ratio data, it's important to use techniques that preserve the ratio between the data points, such as log normalization.

**Why we use Normalization Techniques?**

**To remove the impact of scale:**Data can have vastly different scales and ranges, which can create issues in analysis and modeling. Normalization helps to remove the impact of scale and put all features on the same scale.**To improve performance of models:**Many machine learning algorithms work better when the input data is normalized. Normalizing the data can lead to faster training and better performance of the model.**To address skewness in the data:**Normalization can help to address skewness in the data, which can be caused by outliers or by the data being distributed in a non-normal way. By transforming the data into a more normal distribution, it can be easier to analyze and model.**To improve interpretability of the data:**Normalization can make the data more interpretable and easier to understand. By putting all features on the same scale, it can be easier to see the relationships between different variables and make meaningful comparisons.

**Normalization Techniques:**

**Min-Max Scaling:**This normalization method is utilized to convert information into a range between 0 and 1, by subtracting the minimum value from each data point and after that partitioning by the difference between the greatest and least values. This normalization method is valuable when managing with information that has exceptions, so that the exceptions do not skew the information much as well.**Z-Score Normalization:**This normalization strategy is utilized to convert information into a standard normal distribution, by subtracting the mean from each data point and, after that dividing by the standard deviation. This procedure is valuable when the information contains a normal distribution, because it makes a difference to create the information more interpretable.**Decimal Scaling:**This normalization strategy is utilized to convert information into a range from 0 to 1, by subtracting the minimum value from each data point and after that dividing by the difference between the greatest and least values. This normalization procedure is valuable when managing with exceptionally expansive datasets, because it makes a difference in diminishing the information to a manageable range.**Log Transformation:**This normalization method is utilized to convert information into a logarithmic scale, by taking the log of each data point. This procedure is useful when managing with information that incorporates a wide extend of values, because it makes a difference to decrease the variety in the information. This technique is additionally valuable when managing with information that has outliers, because it makes a difference to decrease their impact on the information.

**Points of interest and Impediments:**

**Advantages:**

- Normalization helps to reduce data redundancy and improve data integrity. By dividing huge tables into smaller, related tables, normalization decreases the amount of information stored in each table. This makes the information less demanding to access and modify, and reduces the sum of capacity space vital.
- Normalization moreover improves information consistency. By organizing data into multiple tables, you can guarantee that the same data isn't stored in different locations. When changes are made, they are reflected in all related tables.
- Normalization also helps to reduce the complexity of queries. By dividing huge tables into smaller, related tables, queries can access as it were the information they require and do not have to process unnecessary data.

**Drawbacks:**

- Normalization can make performance issues. Joining numerous tables together can cause the database to run slower, particularly when a huge number of rows are included.
- Normalization can moreover make it more difficult to query the data. For illustration, in the event that you need to recover information from different tables, you may need to type in complex SQL queries.

Normalization techniques ought to be utilized when data redundancy and data integrity could be a concern. It can moreover be utilized when information consistency is important and when queries need to be simplified. Finally, normalization can offer assistance to reduce the complexity of data structures.

**Normalization in Machine Learning:**

- Normalization in machine learning plays a really important part within the precision of calculations. Normalization makes a difference to scale features so that the information is inside a certain range, usually between 0 and 1. This guarantees that all features contribute equally to the analysis, otherwise it might lead to bias towards one include.
- Normalization too helps in increasing the convergence rate of machine learning algorithms such as clustering, neural networks, and regression. Typically since the algorithms work better when the data points are near to each other and inside the same range. With normalization, the data points are more homogenous and the machine learning algorithm can learn and make more accurate predictions
- Additionally, normalization makes a difference to decrease the sum of noise in the data. When the information is centered around a mean of zero, it'll be simpler to identify the vital designs and relationships. This will lead to better comes about and more exact predictions.

**Implementing Normalization:**

Lets consider the **A_Z Handwritten Data.csv** to demonstrate data normalization in python

Link: https://www.kaggle.com/datasets/sachinpatel21/az-handwritten-alphabets-in-csv-format

```
import pandas as pd
import numpy as np
from sklearn.preprocessing import StandardScaler
# read the dataset
data = pd.read_csv('handwritten_recognition_data.csv')
# separate the features and labels
X = data.iloc[:,:-1]
y = data.iloc[:,-1]
# normalize the features (scale them to have zero mean and unit variance)
scaler = StandardScaler()
X_normalized = scaler.fit_transform(X)
# convert the normalized features to a DataFrame
X_normalized_df = pd.DataFrame(X_normalized, columns=X.columns)
# print the normalized features
print(X_normalized_df)
```

This code demonstrates how to implement normalization using Python libraries such as Pandas, NumPy, and Scikit-Learn. The code reads in a handwritten recognition dataset, separates the features and labels, and normalizes the features using the StandardScaler from Scikit-Learn. The normalized features are then converted to a DataFrame which can be printed to verify that the normalization has been applied correctly.

**Real-world examples of normalization in data science**

**Customer Segmentation**

An online retail company is using customer segmentation to better understand its customer base and target potential new customers. The company collects data on customer demographics, purchasing habits, and product preferences. In order to segment their customers, they normalize this data by taking into account a variety of factors such as age, location, income, gender, and purchase frequency. This normalization of data allows the company to group customers into similar segments, which they can then use to target marketing campaigns or create personalized product recommendations.

**Image Recognition**

An artificial intelligence company is developing a system that can identify and classify different types of images. In order to do this, they first normalize the images by adjusting the brightness, contrast, and colour saturation to ensure that the images are consistent and can be accurately classified. Then, they use machine learning algorithms to identify objects within the images and classify them accordingly. This process of normalizing the images and using machine learning algorithms allows the system to accurately identify and classify different types of images.

**Fraud Detection**

A financial institution is using data normalization to detect fraudulent transactions. They collect data on customer transaction patterns, such as frequency and amount of transactions. By normalizing this data, they can identify outliers or suspicious patterns that may indicate fraudulent activity. The normalized data can then be used to trigger automated alerts or further investigation into potentially fraudulent transactions. This process of normalizing data allows the financial institution to detect and prevent fraud.

**Conclusion**

After the data were normalized, the company was able to use the data to gain meaningful insights about its customer base and make better-informed decisions about its marketing and sales strategies. The company was also able to use the data to create more accurate and powerful algorithms to power its analytics and machine learning processes.

**Key takeaways**

- Use a standard normalization technique, such as min-max or z-score, to transform your data into a common range.
- Ensure your data is in a proper format to begin with by using an appropriate data type.
- Use built-in Python modules such as Numpy and Pandas for efficient data normalization.
- Avoid data leakage by splitting your data into training and test sets before normalizing it.
- Choose the right normalization technique based on the data distribution and the desired outcome.
- Re-check your data after normalization to ensure it is within the desired range.

**Quiz**

**What is the main purpose of normalization in machine learning?**- To reduce the complexity of data
- To make data more interpretable
- To reduce the variance of data
- To reduce the risk of overfitting

**Answer**:c. To reduce the variance of data

**What is the most common type of normalization used in machine learning?**- Z-Score Normalization
- Min-Max Normalization
- Decimal Scaling
- Feature Scaling

**Answer**:b. Min-Max Normalization

**What is the range of values for data normalized using min-max normalization?**- 0 to 1
- -1 to 1
- 0 to 100
- -100 to 100

**Answer**:a. 0 to 1

**What is a benefit of using data normalization?**- It can speed up the training process
- It can help make data more interpretable
- It can reduce the complexity of data
- It can reduce the risk of overfitting

**Answer**:a. It can speed up the training process

Made with

in Bengaluru, India - Join AlmaBetter
- Sign Up
- Become A Coach
- Coach Login

- Contact Us
- admissions@almabetter.com
- 08046008400

- Location
- 4th floor, 133/2, Janardhan Towers, Residency Road, Bengaluru, Karnataka, 560025

- Follow Us

© 2022 AlmaBetter