Overview

This system employs machine learning to detect planets by analyzing specific parameters in observed astronomical data. Through meticulous parameter selection and analysis, the model uncovers potential signals indicative of exoplanetary presence. The system's success contributes to the expansion of our knowledge of celestial bodies beyond our solar system

Disclaimer

This application has been designed for demonstration and preview purposes. While the results generated by the model may not fully align with real-world scenarios, this service is intended to assess the model's fundamental capabilities and potential. In the coming phases, precision will be exercised in refining the model, and adjustments will be made based on feedback received from the demonstration.

Interface Guide

Left Bar: In the left bar, you can read the parameter's explanation

Sliders: Sliders are the adjustable features

Spotify Playlist: Have you ever pondered the auditory essence of a planet? If so, then welcome to this playlist!

Demo Adjustable Parameters Explanation

In our demo project, we made a clear documentation to understand the parameters of the project more properly

Google Drive Link

Thank you to my friend for preparing this unique documentation for the project! Here is the Github of my friend!

Models and Their Explanations

As a basic demo model, we've been using a Random Forest Classifier and at least 1000 rows and 9 columns which are called demo model.joblib. This model is just only to display the project in Streamlit -you may see that in above- so that's why we ignored get valuable accuracy in the model.

There is also expanded demo model that has more capacity to estimate planet based on the parameters given by the user. This expanded demo model has the ability to give you more accuracy (between %50 to %56) and we used almost 9000 rows and 9 columns to make it more complex. The expanded model name is model.joblib

Model's Driver

You can access the models by using the link

Why Random Forest Classifier?

Machine Learning Algorithms

Algorithm	Accuracy
Decision Tree Classifier	38%
Decision Tree Classifier (RandomizedSearchCV)	22%
Logistic Regression	20%
SVC	19%
KNeighborClassifier	16%
KNeighborClassifier (With Best Params)	19%

Deep Learning Algorithm

Keras Sequential with PCA

20%

What is the negative impact for the models?

In demo process, we detected that the dataset is not enough to explore complex relationships among the features so that's why the accuracy of each model are not fitting with our project's request. First of all, I will collect more proper data but the major thing is Feature Selection results, I'll collect the data based on the feature selection results.

The another issue is using resources. In the end of the day, more data will want more resources, the plan is about using Principal Component Analysis

Resource Using

In this demo project, we used one of the main concept of Joblib, that's Memory Cache. It is cache memory, also called cache, supplementary memory system that temporarily stores frequently used instructions and data for quicker processing by the central processing unit (CPU) of a computer. The cache augments, and is an extension of, a computer's main memory.

Demo Delights: Showcasing My Top Features and Functionalities

OOP schema

Cosine Similarity

ML&DL algorithms

Streamlit

Python

Dataset

Dataset is provided by NASA Exoplanet Archive which is completely free to use. Here is the link

How to use?

Write pip install requirements.txt to CMD(or relevant)

Write streamlit run app.py when you are in src/Planets/components to CMD(or relevant)