Machine learning algorithms | Introduction to machine learning

Types of machine learning

1. Supervised learning: In a supervised learning model, the algorithm learns on a labeled data set to generate expected predictions for the solution to new data.

P.ej; For house price forecasting, first we need data on dwellings such as; square foot, no. of rooms, the house has a garden or not, and so on. So we need to know the prices of these houses, In other words; class labels. Now that the data comes from hundreds of houses, its characteristics and prices, we can now train a supervised machine learning model to predict the price of a new home based on past experiences of the model.

Supervised learning is of two types:

a) Classification: In Classification, a computer program trains on a training data set and, according to training, categorize data into different class labels. This algorithm is used to predict the discrete values ​​as male | woman, true | fake, spam | no spam, etc.

P.ej; Spam detection, voice accreditation, cancer cell identification, etc.

Types of classification algorithms:

• Naive Bayes classifier
• Decision trees
• Logistic regression
• K-Nearest Neighbors
• Support vector machines
• Random forest classification

b) Regression: The task of the regression algorithm is to find the mapping function to map the input variables (x) to continuous output variable (Y). Regression algorithms are used to predict continuous values ​​such as price, salary, age, grades, etc.

P.ej; Time prediction, house price prediction, fake news detection, etc.

Types of regression algorithms:

• Simple linear regression
• Multiple linear regression
• polynomial regression
• Decision tree regression
• Random forest regression
• Set method

2. Unsupervised learning: In an unsupervised learning model, the algorithm learns on an unlabeled data set and tries to make sense by extracting features, co-occurrence and underlying patterns alone.

P.ej; Anomaly detection, including fraud detection. Another example is the opening of emergency hospitals to areas with the highest risk of accidents.. Clustering K-means will group these maximum prone area locations into clusters and define a cluster center (In other words, hospital) for each group (In other words, accident prone areas).

Types of unsupervised learning:

• Grouping
• Anomaly detection
• Association
• Car coders
• Latent variable models
• Neural networks

3. Reinforcement learning: Reinforcement learning is a type of machine learning in which the model learns to behave in an environment by performing some actions and analyzing the reactions. RL takes the appropriate measures to maximize the positive solution in the particular situation. The reinforcement model decides what actions to take to perform a given task, that is why he is obliged to learn from experience itself.

P.ej; Let's take an example of a baby learning to walk. In the first case, when the baby starts to walk and reaches the chocolate given that chocolate is the baby's end goal and a baby's solution is positive given that he is happy. In the second case, when the baby starts to walk and while walking is hit by the chair and cannot get the chocolate, starts crying, which is a negative answer. In other words, how humans learn from trace and error. Here, the baby is the “agent”, chocolate is the “reward” and many obstacles in between. Now the agent tries in various ways and discovers the best feasible way to achieve the reward.

Machine learning life cycle

Machine learning helps increase task performance and productivity. Includes learning and self-correction when presented with new data.

The machine learning life cycle involves six main steps:

Paso 1: data collection

Identify multiple data sources like Kaggle and collect the required dataset

Paso 2: data pre-processing and EDA

In this step, we do an analysis of the data in search of missing values, duplicate data, invalid data using different analytical techniques. And also preprocess the data for feature extractions, feature analysis and data visualization.

Paso 3: model train

We use a data set to train the model using various machine learning algorithms. Training a model is essential for it to understand the various patterns, rules and characteristics.

Paso 4: test model

In this step, we verify the accuracy of our model by providing a test data set to the trained model.

Paso 5: Model implementation

Model implementation means integrating a machine learning model into an existing production environment that takes inputs and returns results to make data-driven business decisions. Several technologies that you can use to put your machine learning models into practice are listed.:

• Stowage
• Governors
• AWS SageMaker
• MLFlow
• Azure Machine Learning Service

Paso 6: Model monitoring

After model implementation, here comes model monitoring which monitors your machine learning models for factors like errors, failures and latency and, the most important, to ensure your model maintains the desired performance. Model monitoring is very important as your models will degrade over time due to various factors, as invisible data, changes in the environment and relationships between variables.

Some applications of machine learning in the real world

• Automatic language translation in Google Translate
• Faster route selection in the Google Map search engine
• Driverless car / autonomous
• Smartphone with facial accreditation
• Voice accreditation
• Ad recommendation system
• Netflix recommendation system
• Suggestion of automatic tagging of friends on Facebook
• Stock market negotiation
• Fraud detection
• Time prediction
• Medical diagnostic
• Chatbot
• Machine learning in agriculture

Benefits of machine learning

• Work automation
• Powerful predictive ability
• Increased sales in the e-commerce market.
• Benefits of AA in the medical domain to drive medical diagnosis and drug development
• Machine learning is used in robotic medical surgery
• ML in finance increases productivity improves revenue and provides secure transactions
• Model the data to make useful decisions

Summary

Machine learning can be used in almost every sector of human life to do our work efficient, robust, Y Without complications. As we know, everything has its pros and cons, machine learning also has its downsides, as an example, with the rise of machine learning, many people may lose their current stage job. But more in a bombastic tone is beneficial in the long run for humanity.