Exoplanet Identifier

A Machine Learning Project to Classify Exoplanets

Why am I doing this?

I say I understand machine learning (which I do), but I haven't really gotten my hands dirty.

I have almost no experience with tools like pandas, scikit-learn, numpy, etc., but here’s how we learn by jumping into the deep end 🚀
I want to see where I struggle the most, this is also really going to help with ARA

---

What is the Goal Here?

I want to analyze planetary properties to determine if a planet is:

• 0 → "Candidate" → A potential exoplanet
• 1 → "Confirmed" → A verified exoplanet
• 2 → "False Positive" → Not actually an exoplanet

I also want to build a solid ML foundation before diving into Deep Learning (Neural Networks and all that fun stuff) while working on the virtual assistant for ARA.

---

What’s in the Data?

The dataset comes from Kaggle's Kepler Exoplanet Dataset, which includes the following key features:

Feature	Description
kepid	Unique Identifier for the host star
kepoi_name	Unique Identifier for the planetary candidate
koi_disposition	Status of the planetary candidate (0,1,2)
koi_score	Confidence score for classification (higher = stronger confidence)
koi_period	Orbital period (in days)
koi_prad	Estimated planetary radius (in Earth radii)
koi_teq	Estimated equilibrium temperature (Kelvin)
koi_insol	Insolation flux received (relative to Earth)
koi_steff	Effective temperature of the host star (Kelvin)
koi_srad	Stellar radius (in solar radii)
koi_slogg	Surface gravity of the host star (cm/s²)
koi_kepmag	Kepler-band magnitude (brightness of the star)

---

What’s Next?

1. Load and explore the dataset using pandas 📊
2. Apply feature engineering & preprocessing 🛠
3. Train a machine learning model to classify exoplanets 🤖
4. Evaluate performance and optimize the model 📈

---

Project Roadmap

✅ Step 1: Load the data and check for missing values
✅ Step 2: Perform exploratory data analysis (EDA)
✅ Step 3: Preprocess data (handle missing values, normalize features)
✅ Step 4: Train an initial ML model (Random Forest / XGBoost)
✅ Step 5: Tune hyperparameters and optimize performance
✅ Step 6: Interpret results & visualize feature importance

---

Loading Project Data

I loaded the Exoplanet data from the csv to a dataframe object using pandas read_csv() function I printed the head(top 10 rows of dataframe) to verify the correct data and labeling Now it is time to preprocess the data

Visualize the data

I want to do this before I preprocess.
This can answer these questions: How is the data distributed? How many outliers are there? How skewed is the data?

I also want to see any correlation between the data features

Doing this will help select the normalization technique\

Identify the independent and dependent variables

This requires a good understanding of the dataset which at the moment I thought I did
Granted I took it for face value instead of looking into it which I will once I get home or should I pay attention to my assembly class?\

What kind of correlation do all these features have? What makes a planets an exoplanet according to these features? Ok here is what I found: \

Data Preprocessing

Now it is time to prepare/clean the data
This consists of: 1. Handling missing data 2. Removing Redundant Data 3. Normalizing data

Handling Missing data

After checking for any missing values is used exoplanet_df.is_null().any() to see if and where I am missing data
Doing this revealed that there is no missing data which is great \

Removing Redundant Data

Now it is time to remove redundant data
This is data that has no importance but can affect the models performance
That data in this case is: kepid kepoi_name
These labels and data provide no useful insight and can affect my models performance

Data Normalizing

Now I have to normalize the data
What is Data Normalization?
Data Normalization is transforming data features to a standard scale between 0 and 1
This is done by adjusting features based on maximum and minimum values
You do this to ensure not one feature dominates the others

There are several ways to normalize data
One way I found that is very common is the Min-Max formula
Looks like this:

X_norm = (x - x_min) / (x_max - X_min)

Where:
X is the individual data
X_min is the smallest value in the dataset
X_max is the largest value in the dataset
X_norm is the normalized value(Usually between 0 and 1)\

There are other ways such as Max-Abs Normalization where:
X_norm = X / |X_max|

or

Mean Normalization or Z-Score Normalization But I want to check out Log

The Machine Learning Model

At this point in time I am beginning to think about the model I want to use
First I should probably graph the points or something
For now though I am thinking of using Linear Regression(assuming this data is pretty linear)
Otherwise I will have to look at the data and see where it is going\