DEV Community: kainat Raisa

🔍 Understanding Logistic Regression for Classification

kainat Raisa — Sun, 24 Nov 2024 06:04:32 +0000

Logistic regression is a powerful tool in machine learning, especially when dealing with binary, multinomial, or ordinal classification tasks. Here’s a quick breakdown of how it works and when to use it:

1️⃣ Logistic Regression Basics:

It’s mainly used for binary classification (where the output is either 0 or 1).
The model uses a sigmoid function to output a probability value between 0 and 1.
The logistic curve, or S-shaped curve, predicts values between 0 and 1, giving us a probabilistic interpretation.
2️⃣ Sigmoid Function:

If the sigmoid function output is greater than 0.5, the datapoint is classified as Class 1, otherwise Class 0.

3️⃣ Softmax for Multiclass Classification: For problems with more than two classes, we use the Softmax function to handle multiclass classification:

4️⃣ How Logistic Regression Works:

The model computes a linear combination of the input features:
𝑧 = 𝑤 ⋅ 𝑋 + 𝑏

Where w represents the weights and b is the bias.
Then, it applies the sigmoid function:

This converts the linear output to a probability.
5️⃣ When to Use Logistic Regression:

Best suited for datasets without outliers.
It works well when there’s a clear decision threshold.
If you're diving into machine learning, logistic regression is a must-know! It's a great starting point for understanding classification problems. 🚀

MachineLearning #DataScience #LogisticRegression #AI #Classification #SigmoidFunction #Softmax #BinaryClassification

Space Mission ML Project - Part 2(Data Visualization)

kainat Raisa — Mon, 13 Nov 2023 14:13:43 +0000

Let's Visualize the Story, our Space Mission Dataset is trying to tell

In the first part of this blog have tried to separate the mission details of each organization.

Now we are going to visualize how the cost/price of the space missions of a specific organization have varied/changed over time.

Here we have used the Plotly visualization library to create a date v/s price for ISRO.

Now we'll see the frequency distribution of all mission costs. So we have created a distribution plot here using the matplotlib and seraborn library(python's seaborn library has been built on the roof of matplotlib).

Now we'll see counts of the status of the missions which means how many space missions are in each status worldwide.

Probably a concern is raising in your mind what if we get to see how many total missions have been launched by each space organization?
here we have created a bar plot.

To see the success or failure rate of the missions we can create a piechart. When we are creating a piechart with the Mission-status and count column(created using value-counts) using plotly the counts will default converted to percentages.

Understanding Machine learning

kainat Raisa — Sun, 12 Nov 2023 08:30:52 +0000

How do humans learn?

Understanding and explaining how humans learn is a complex discussion as it involves Psychology intensively. But if I have to explain "how humans learn" in simple words I will say that when our brain gets exposed to Data or when we encounter a situation our brain explores the situation and trains itself using the data. When we encounter a similar situation our brain makes decisions or acts based on the pre-training. Now moving forward, why am I talking about the process and psychology behind how humans learn? Machine Learning or how a machine learns has a lot of similarities with the human learning process.

Converting NASA's Open Data into Insights using EONET API

kainat Raisa — Mon, 23 Oct 2023 13:29:59 +0000

About the Project

In this project, I have tried to convert the NASA EONET Open Data into a bit more visualized form. This is a very small project where I have accessed the Data about Earth's Natural Events using the Open EONET API from NASA and created an interactive world map.

Importing the necessary Python Libraries

Here I am importing the requests, json, pandas, folium, IPython Library of Python. If you are using Jupyter Notebook and don't have the libraries preinstalled on your desktop then just write the code in a cell below (separately for each library) :

Here I have put the API URL and URL key in two variables named as url and key, and then I have set the parameters for making the API request ( if you want to know more about the NASA Open APIs then click on this link: https://api.nasa.gov/ )

Making the API request and printing the textual part of the response

The response is returning us a JSON object (a javascript object), this format of data is almost similar to a Python dictionary.

Now we'll load the data from the json object.

Converting the title, description, and link part of the response Data into a pandas Dataframe. And then I'll check the head of the Dataframe.

As you can see all of the are in a single column in our dataframe which means they are not in the normal form so we need to normalize the dataframe.

We have normalized the dataframe using json_normalize method of the pandas dataframe.

Now we'll need to separate the event ids and category titles from the categories column.

Now we'll see how many of the different natural events happening on earth on that particular data.

Now I am using the folium library to create a world map.

I have saved the map on my desktop as an HTML file as world_natural_events_map2.html. Now I'll use the IPython library to display the map.

this is an interactive map, we can click on the markers and see further details about the events.

Space Mission ML Project - Part 1(EDA)

kainat Raisa — Thu, 19 Oct 2023 07:15:31 +0000

About the Dataset

The Kaggle Dataset today we'll be using for this machine learning project is the space mission launch dataset. This dataset is about the status, cost, rocket status, etc. of the Space missions launched by different space organizations over time.
You can download the dataset from the link below.
Link of the Kaggle Dataset: https://www.kaggle.com/datasets/sefercanapaydn/mission-launches

About the Project

The Machine Learning Project we are going to create today is about exploring how the space mission cost has changed over time, which space organization has seen the most success or failures over time etc., and will create a predictive model to predict the success/failure of the future space missions. We are going to use some amazingly useful Python libraries and Python to create meaningful stories from the data.

Unleashing the Magic from the Data Begins

Importing the Essential Libraries

In this part of the project, we'll be importing just the libraries we are going to use for the data analysis ( the ML libraries will be introduced in the 3rd part of the project).

( don't freak out right now, I'll explain everything about the libraries and when to use which one )

Accessing the Data

Here we have read the CSV Data file and restructured the data into a pandas DataFrame. Using the head() method we are trying to see the first 5 rows/entries of the dataset. This shows us what and how the datapoints are present in the dataframe.

Cleansing and Filtering the Data

We are dropping the two unnamed columns using the drop(inplace=True) (inplace=True represents that we are dropping the columns from the original dataframe) method as they are not providing much value to the dataset. Then we'll check the head() of the dataframe again just to ensure that the columns are no longer in the dataframe.

The info() method returns the non-null datapoints account and types of the datapoints.

the isna() method shows us whether there are null values or not. We usually drop the null values when there are very few of them but in this case, the Price column is carrying a large amount of null values so we'll handle the null entries differently.

here we can see from the info that the price column is carrying non-numeric values so we are converting the prices into numeric data.

Let's see what the Data is telling us

We have organized details about all of the missions each organization has launched. Let's see the first 10 entries here:

Here we are trying to see total number of missions of each organization and the organization name with the maximum missions :

Now we'll see how many missions have been successful, how many have failed, and how many have prelaunch and partial failures for each organization.

As we know a space mission's success or failure is a significant parameter in an organization's space research history so we need to know the name of the organization with the maximum successful missions.

To know more and all of the details about an organization we are trying to filter out the data by organizations. Enter a space organization's name and their space mission details will be shown. (here we have printed the Data for ISRO. Go and try it out for SpaceX)

Here we are trying to know the status of the rockets of each organization.

Machine Learning-The modern way of unleashing hidden possibilities

kainat Raisa — Wed, 19 Apr 2023 08:17:36 +0000

Why Machine Learning ?

I always love to start my blogs with explaining "why" we are even talking about that topic. The reason behind this is that i strongly believe that the "why" should always be clear to us whenever we are heading towards something new. When we have the answer of "what is it?" we basically just "know" the topic but when we have the answer of "why this?" we start understanding the topic. You are maybe thinking that why am i discussing all of these philosophical stuff but these words are for those who will become learners from just readers of this blog.

Getting back to the point, now we will discuss that why Machine Learning had been introduced to human. In typical words Machine Learning is the ability of computational machines to imitate human-like behavior/intelligence. What type of human intelligence we are talking referring to? We are talking about human's ability of recognizing hidden patterns and predicting things. We can teach a machine how to unleash hidden patterns and make predictions based on the patterns. This is basic concept of Machine Learning. Our traditional computation system does not allow the machines to learn from given data by their own but ML allows us to do so. Application of ML in various fields like business, Education, Research etc. has unleashed many hidden possibilities.

Machine Learning Categories

There are three ways of a machine to learn from data, discover the hidden patterns and unleash the story behind the dataset. When we train a machine learning model using labelled data which means we tell the machine what the input dataset is about and what to predict is called Supervised Learning. The second way of training an ML model is using un-labelled dataset which means we won't tell the model what is the input dataset about and what to predict. Is is sounding absurd? okay, let me explain this a bit more. Suppose we are giving the medical history of the cardiac patients a doctor has seen in a year as the input dataset to train an ML model. We are not telling the model what we actually want the model to predict. Suppose maximum of the patients who were detected to have severe cardiac issue had diabetes, so if we give the medical history of a new patient who visited to the cardiologist the ML model will be able to predict that whether that person has a severe cardiac issue or not before the physical check-up.
The third way an ML model can be trained is through reward or punishment based learning which is called Reinforcement Learning. In this method the model will get rewarded(in the form of positive feedback) for producing the desired behaviors and will get punished(in the form of negative feedback) for undesired behaviors.

Linked Lists (implementation using Python) and it’s applications

kainat Raisa — Mon, 27 Mar 2023 14:58:01 +0000

Let's talk about Lists first

In computer programming Lists are a type of data structure through which we store data. List data structures are similar to the "real life list" concept. The way all of the elements of a real life list comes one after another list data structures also store the data contiguously in the memory. We can access, modify as well as delete the elements of a list data structure.

 (The theoretical part)

What are Linked Lists

Linked Lists are the data structures which store data linearly (like a list) but they don’t store the elements contiguously in the memory. Each element of a Linked List is called a Node. The links among the nodes is a unique feature of a Linked List. Based on the connections there are 4 types of Linked Lists. The 4 types of Linked Lists are: Singly Linked List, Circular Singly Linked List, Doubly Linked List and Circular Doubly Linked List.
Enough of the bookish definition of Linked Lists now we'll learn about Linked Lists in our own words.
Every linked list has atleast concepts/parts. A head node/value, a tail node/value, next and pervious node's reference.

🔶 Singly Linked List:
Let's assume the Singly Linked list as a train. The way the train compartments are connected to their next compartments with a link, the nodes of a Singly Linked List are linked to their next nodes. Every element of a Linked List is called a node. The very first node of the Linked List is called the Head. The last node is called the tail(the node after which no newnode will exist is the last node). As we know that Linked Lists don’t store data sequentially so index 1 stores the tail node(this is the 2nd node we create). A node stores the memory address of the next node. We can access the next node of a linked list using the node.

Real life application: Back buttons(for example:- the smartphone back buttons are created using Singly Linked List so they allow us to go only in a single direction).

🔶 Circular Singly Linked List:
The only structural difference between Singly Linked List and Circular Singly Linked List is that the Circular Singly Linked List has a link between the Head and Tail node. Which means the Tail node stores the memory value of the Head node.

Real life application: The Circular Singly Linked List is used in our personal computers. When multiple applications are run on a PC all of the running apps are stored in a CSLL.

🔶 Doubly Linked List:
The nodes of a Doubly Linked List store the memory locations of both the previous and next node. Which means we can access both the previous and next node using the current node. Otherwise all of the features of DLL are similar to SLL.

Real life applications: Web browser cache, music player etc.

🔶 Circular Doubly Linked list:
The only structural difference between Circular Singly Linked List and Circular Doubly Linked List is that in CDLL the Head and Tail nodes are connected with eachother both ways so the tail's next node will be the head and the head's previous node will be the tail node.

Real life applications: Shopping cart on online websites etc.

    (The implementation part will be added here soon)

Notes on Data Science Fundamentals

kainat Raisa — Sun, 05 Mar 2023 15:25:42 +0000

Why Data Science?
Huge amount of data is produced everyday from virtual platforms, industrial zones, corporate offices, Scientific experiments etc. All of these data needs to be managed, organized, analyzed properly to be used for making better strategies for our businesses, organizations, research work etc. Data is a core part of the development as well as the revolution of 21st century. So we need a whole field where we will work,understand, discover with data and from data. This field of work and study is know as Data Science.

What is Data Science?
The science of getting insights from data(usually huge amount of data) by collecting, organizing summarizing, and visualizing them for better decision making is called Data Science.

Some Tools we use for Data Science :-

▪️Databases (MySQL, MongoDB etc.) : Used for storing data in a structured way.
▪️Spreadsheet softwares (MS Excel, Google Drive Spreadsheet etc.) : Used to prepare Data for analysis.
▪️ Data analysis tools (Python, R, MATLAB etc.) : Used for structure, analyze, visualize data.
▪️ Programming libraries and modules (Matplotlib, seaborn, plotly, Numpy,Pandas etc.) : Used for performing different operations on the Data.
▪️ Apache Spark : Used for big data loads.
▪️ Virtual Environments ( Jupyter Notebook, VSCode etc.) : Used to perform the operations on data using programming languages.

(There are many more tools which are widely used in the field of Data Science)

Data Science methodology( steps we follow for Data Science) :-
▪️Business Understanding
▪️Analytic Approach
▪️Data requirements
▪️Data Collection
▪️Data understanding
▪️Data preparation
▪️Modeling
▪️Evaluation
▪️Deployment
▪️Feedback

(All of the concepts mentioned above will be broadly explained in another blog)

Time Complexity-Diving into the ocean of Algorithms(part 1)

kainat Raisa — Wed, 01 Mar 2023 11:19:43 +0000

What do we mean by Time Complexity?

I am assuming that the person who is reading this blog is just like me who always freaks out whenever it comes on understanding Data Structures and Algorithms. Don't worry mate, in this blog we are going to figure out together how to deal this "Problem".

Let's understand how a computer programming algorithm works. We can compare a computer(any computing machine) with a human brain. In our day to day life our brain follows multiple sequential commands to accomplish a particular task. Just like that a computer follows the sets of instructions we as a computer programmers give to it to perform any task. The set of instructions/steps a computer follows to accomplish a task is called Algorithm. A computer takes time to execute every programmig algorithm. But the time requirement for every algorithm is not the same. Different algorithms take different amount of time to get executed(just like making tea and doing your assignment requires you different amount of times). On the other hand the execution time of an algorithm varies with the input size(the amount of material we work with). For example: The time you’ll need/spend to make tea for ten people will be more than the time you’ll need to make one cup of tea,because you’ll need more water/milk for ten cups of tea than a single cup of tea and more amount of water/milk requires more time and heat to boil. Just like this when you write an algorithm of finding sum of "n"(n is a variable which can be any positive integer)numbers the time to execute this algorithm for two numbers will be less than the time required for 100 numbers.
Again, when we are talking about efficiency of performing a task we can not forget to consider the capability of the performer. For example: Your mother takes less time to make tea than you spend to make(as you need to find everything and end up making a mess in the kitchen 😐). This is because your mother is more efficient in executing the tea making process but you are a confused kiddo! Just like this different devices take different amount of times to execute a program.

The time variation of an algorithm due to the changing input size is called "Time Complexity of an Algorithm".

How do we measure Time Complexity of an Algorithm

As we have already discussed that the execution time of a computer program very easy with the input size as well as the device we run the program on. So it will not be appropriate for us two measure the absolute valuevalue of the execution time. So the mathematicians have come up with the idea of expressing the execution time as a relationship between the input size and time.
This relationship is expressed using different equations for different types of algorithms.The relationship between the input size and the time required can be linear or nonlinear. For example if we loop through a list Data structure using a single loop then the relationship between the input size and execution time can be expressed like a linear funtion.
There are three cases(we will consider these three) of time complexity. These three cases are expressed with three notations which are called the Asymptotic Notations. When the program works with the minimum input size possible and takes minimum possible time to be executed we call it the Best case time complexity and it is expressed with Big Omega(Ω(n))notation. When our program has to work with the maximum possible input size it requires maximum time to execute this case is the Worst case time complexity and it is expressed with Big O (O(n)) notation. The Average case time complexity is the average of Big O and Big Omega complexity which is expressed with Big Theta(θ(n)) notation.

Often a programming problem is solved using multiple algorithms
but not every algorithm is time efficient. When we talk about time efficiency of a computer program we actually compare the worst case time complexity which means we consider Big O.
Big O or the worst case time complexity expression comes from the relationship of time and input size we talked about.

pandas, a Python library for data manipulation

kainat Raisa — Sun, 12 Feb 2023 07:20:34 +0000

How can a panda manipulate data?

When i first heard the name 'pandas' i wondered "how can these chubby animals work with data?". okay...i know that you are thinking how dumb i was. So now, if in Python 'pandas' are not some chubby entities then what is it???
pandas is a Python library which helps us to work with data. As we know that a code library is a collection of modules or functionalities which can be used by any programmer. Python is one of the languages which provide a huge variety of code libraries. pandas is one of the most helpful and important feature of Python which has made data manipulation way much easier for the tech-people who has to work with bulky datasets.

Why pandas?

Fields like Machine Learning, Artificial Intelligence, Deep Learning, Natural Language Processing, Data Science require a huge amount of data. These data can either be raw or organized data. pandas help us to visually represent the data in a comprehensible format.

Some fundamentals of pandas

Using pandas we can present a data set in a tabular format where all of the data are presented in rows and columns with corresponding row and column headings. In pandas this tabular format of data is called a DataFrame. We can pass CSV files,excel files or even dictionaries as a pandas Dataframe.
We can import pandas just like other python libraries with the import keyword. Here i am inporting pandas as pd( the most convenient abbreviation for pandas).

When we pass a dictionary inside a pandas DataFrame the keys correspond to the column names and values are the column elements.

Now we are going to pass the dictionary named "dict" inside a DataFrame and store the DataFrame in a variable named "df" (you can give any name to the variable). And we'll print the DataFrame df to see the output.

Yeeeyyyy! We have successfully created our first DataFrame.

Entering the world of Object Oriented Programming(OOP)

kainat Raisa — Thu, 08 Dec 2022 06:09:21 +0000

Why OOP?

Before learning anything we should always find the answer of "Why is this useful/Why are we learning this?". So before knowing what is OOP we are going to know why we are even talking about this.
In the physical world we need to create products which are similar to eachother. The products may have the same functionalities but probably with a little different specifications. As we need to use the same base product everytime we work smartly. We make a base skeleton of the products with the similar functionalities and add the specifications later on while creating a particular object. Let's understand this with an example: Suppose you want to build your own house. So what would you do first? You'll hire an architect to make a blueprint of your house. This blueprint is the class which can be called " House" class. This "House" will/may have windows,doors,rooms,balconies and many more things. All of these features are the functionalities of the house. The blueprint of your house can be used by anyone for building their own house. All of the houses which have been built or will be built are "Objects".

What is OOP?

Just like the physical world we need to create many objects with the same functionalities in the virtual world also. So we create a virtual skeleton for that purpose. This is similar to the Resume templates we use to make our own Resumes.In programming languages we use and create many classes and objects to build virtual products. For example: Suppose we are building an "Animal Planet" comouter game which has multiple animals and they all speak different words. In this case "Animal" will be the class and different animals like Dog,Duck and Cat will be the objects. All of them have the "speak" functionality but they speak "Woof"," Quack" and "Meaw correspondingly. In computer programming all of these together(classes, objects, methods etc.) are called Object Oriented Programming. The functionalities are called " methods" in OOP.

The Numpy

kainat Raisa — Mon, 21 Nov 2022 17:33:49 +0000

What we do with Numpy?

The biggest reason behind the popularity of Python programming language is the huge number of built-in libraries it has. Now the question comes to your mind, what is a library? In computer programming a library refers to a collection of some pre-written functionalities which can be reused in any Python code. Python has around 137000 built-in libraries. Numpy(can be written also as NumPy or numpy)is one of the most popular library of Python. Python programmers mainly use this library to work with array,linear algebra, fourier transform and many more mathematical operations.

How to use Numpy

Before using, the Numpy library needs to be installed in the virtual environment we are working in. You can install numpy just by writing "pip install numpy" in the terminal/console.
Then you have to import numpy by writing "import numpy" on the top of your Python code.

Or you can import this library as another name(programmers are too lazy to type this five letter word so they import numpy as np).

How to work with Arrays using Numpy

An array is a Data structure where large number of data can be stored inside a single named space. Arrays can be both one and multidimensional. We can create and panipulate arrays using Numpy library. "numpy.array()" or "np.array()" can be used to create an array.

This is a one dimensional array. A one dimensional array can be considered as a Python list. You can do indexing,slicing, and assign a new value to the one dimensional array just the way you do it with a list.

You can check the shape(how many rows and columns are there), the size(how many elements does the array have) or reshape a multinational array with the shape,size attribute and reshape() method.

random.choice and random.randint

The random.randint(lower limit,upper limit) is used(random and randint are methods) to generate random integers between the given range.

On the otherhand random.choice(sequence)
generates a random value from a given sequence(list,tuple,string etc.).