Learn AI

example 1: Predicting sale value based on the investment in advertisement using linear regression. #ai #agentzai #ml

Objective:- To predict the next day sale values based on the past 200 days of sales data.

Dataset description:- The dataset has four columns ["TV","Radio","Newspaper","sales"] in which first three are to be considered as inputs and one as output.

1. TV:- Daily investment amount (USD) in the TV advertisement.
2. Radio:- Daily investment amount (USD) in the radio advertisement.
3. Newspaper:- Daily investment amount (USD) in a newspaper advertisement.
4. Sales:- End of the day sale values based on the investment.

Usecase:- One of the sources for getting more customers is an advertisement for businesses like Vasanth & co, Sathya like that. The amount they invest in the different sources of advertisements should get them traction. This prediction helps them on which proportion they should invest in TV, radio, and newspaper to get better profit.

Technical Background:- Machine learning has three major categories when it comes to learning the data i.e, supervised learning, unsupervised learning, 0and reinforcement learning. The first two can be thought of as “learning paradigms” and also widely used in building a Machine Learning model. These categories are depending highly on the application and the type of data set available. Labeled data is always beneficial since it can be used in both supervised and unsupervised applications.

a) Supervised Learning: Supervised Learning in which you know the labels and you feed those labeled data into the Machine Learning algorithms for training.

Examples: forecasting weather, forecasting the sales, and classifying objects

Supervised learning is again fall into two categories Regression and Classification.

Regression:- Regression is a target prediction based model which understands the dependency between the independent variables and dependent variables by fitting a line into the data points. In simple words, continuous changes in the input and continuous changes in the output.

example:- weather forecasting, sales forecasting, and stock price forecasting

algorithms used:- Linear regression, lasso regression, and Huber regression.

Classification:- Classification is the process of predicting the label or class of given data. Classes are also called as targets/ labels or categories.

example:- face recognition, object identification , and intent classifications

algorithms used:- Logistic regression, SVM and neural networks.

b) Unsupervised Learning: The difference between supervised and unsupervised learning is the absence of labels/target. The data points are feed into the machine learning algorithm to segregate the datasets into different groups based on the data features. The most important subcategory is clustering which is used to identify the targetted group for a given product/services.

example:- customer segmentation and targetted marketing.

Algorithms used:- K-Means, canopy clustering