When Dr. Ng talked about neural networks, I kinda remembered factor analysis, I know there’s a difference can someone explain it to me?

I also don’t know anything about “factor analysis”, but I googled it and got this description:

**—> quote <—**

**Factor analysis** is a technique that is used to reduce a large number of variables into fewer numbers of factors. This technique extracts maximum common variance from all variables and puts them into a common score. As an index of all variables, we can use this score for further analysis. Factor analysis is part of general linear model (GLM) and this method also assumes several assumptions: there is linear relationship, there is no multicollinearity, it includes relevant variables into analysis, and there is true correlation between variables and factors. Several methods are available, but principal component analysis is used most commonly.

**—> end quote <—**

Based on that description, my conclusion is that factor analysis is a ‘cut and dried’ method (or methods) that can take inputs and produce output based on a mathematical function. For example Principle Component Analysis, which I *have* learned about in Prof Ng’s courses, is a way to use linear algebra techniques to transform the inputs in a way that derives which directions in the input space contain the most information and which directions are not meaningful and can be ignored for efficiency.

What Neural Networks do is not simply a mathematical formula. It is a way to construct a sequence of computational layers that are “programmable” and then a way to learn how to adjust the parameters of each layer so that the whole network solves a given problem, as measured by a cost function. So it is not “cut and dried” in the same way that PCA is: you learn the solution by measuring and then adjusting the performance based on your training data.

I totally grant you that is not a complete answer to your question. But my suggestion is that you start with this thought: factor analysis and neural networks are not the same thing, even though you could view the overall objective as similar in that both start with a lot of information and reduce it to less but more specific information. But the way each method accomplishes that goal is completely different. Then with that thought in mind, continue with the course and listen to all that Prof Ng explains. Once you have digested that, if you actually know about factor analysis, it should be more clear to you how they are different and you can come back and explain the differences to us in more detail.

Hello Khushal,

Taking Paul sir’s explanation on factor analysis to another level, let me take you through the following example:

Suppose you are into a marketing team, which would like to understand the preferences and motivations of their target customer base having data on various factors such as:

1- What products do customers buy?

2- How do users navigate the website?

3- Demographic understanding on Age, gender, income, etc.

4. Survey opinions on product features and brand image.

Using Factor analysis technique here, you can come up with two main factors:

a) Price-conscious value seekers

b) Trendy brand enthusiasts

Right?

The first factor retrieves variables like frequent purchase of discounted items and sensitivity to price changes, while the second factor relates on browsing fashion trends, engaging with social media campaigns, and valuing brand reputation.

This would breakdown your research into 1. Market segmentation by identifying distinct customer groups having unique preferences to target them effectively with personalized messaging and marketing campaigns. 2. Product development analysis by prioritizing features and product offerings that resonate with the different customer segments & last but not the least 3. Brand positioning that tailors brand messaging and communications to appeal to the specific motivations of each segment.

Subsequently, Neural Network technique would learn the purpose to understand complex relationships and patterns from data to make predictions or classifications; Applications used in a wide range of tasks, including image recognition, natural language processing, and time series forecasting; To identify interconnected layers of neurons that process information and learn from training data; Benefits to handle non-linear relationships, learn complex representations of data, and achieve high accuracy in predictions; and look for computationally expensive, prone to overfitting, and difficult to interpret due to the “black box” nature of their internal workings.

Moreover, combination of both factor analysis and neural networks techniques could bring flying colors in learning or training a model using Preprocessing, Feature engineering and Regularization techniques to a model.

**I hope, you get it what I have tried to explain above after a profound search regarding factor analysis through Google**.