https://github.com/samia35-2973/living-type-classification-from-codon-usage
Machine learning project to classify living types based on codon usage data using Random Forest and XGBoost classifiers.
https://github.com/samia35-2973/living-type-classification-from-codon-usage
classification codon-usage data-cleaning data-preprocessing excel exploratory-data-analysis living-type machine-learning python random-forest-classifier scikit-learn supervised-learning xgboost-classifier
Last synced: 7 months ago
JSON representation
Machine learning project to classify living types based on codon usage data using Random Forest and XGBoost classifiers.
- Host: GitHub
- URL: https://github.com/samia35-2973/living-type-classification-from-codon-usage
- Owner: Samia35-2973
- Created: 2025-02-28T03:50:45.000Z (7 months ago)
- Default Branch: main
- Last Pushed: 2025-02-28T03:57:11.000Z (7 months ago)
- Last Synced: 2025-02-28T11:26:57.839Z (7 months ago)
- Topics: classification, codon-usage, data-cleaning, data-preprocessing, excel, exploratory-data-analysis, living-type, machine-learning, python, random-forest-classifier, scikit-learn, supervised-learning, xgboost-classifier
- Language: Jupyter Notebook
- Homepage:
- Size: 8.73 MB
- Stars: 0
- Watchers: 1
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# Living Type Classification from Codon Usage
In this project, I participated in a hackathon where I worked with a raw codon usage dataset. The goal was to classify the data into different living types by cleaning the data thoroughly. The classes of the 'type' were:
- bacteria
- virus
- plant
- bacteriophage
- mammal
- vertebrate
- invertebrate
- rodent
- archaea
- primate
- plasmid
## Directory Structure
```
- codon-usage-dataset-partially-cleaned
- dna_test - dna_test.csv.csv
- train_replaced_missing_values.csv
- Raw Data
- dna_sample_submission.csv
- dna_test.csv
- dna_train.csv
- output
- submission3.csv
- submission4.csv
- submission5.csv
- living-type-classification-from-codon-usage.ipynb
```
### Data Cleaning
1. **Handling Missing Values**: The first step was to load the raw data (`dna_train.csv` and `dna_test.csv`) into Excel, where I addressed string values indicating `NaN`. These string values caused the column's data type to be recognized as `Object`. Once cleaned, the columns were automatically formatted with the correct data types. The cleaned datasets were saved as:
- `train_replaced_missing_values.csv`
- `dna_test - dna_test.csv.csv`
2. **Exploratory Data Analysis (EDA)**: After inspecting the data, I performed EDA to identify and impute remaining missing values. I used data distribution analysis to impute numerical variables. During this analysis, I identified an issue with the `uuu` column, which contained a string value among the float values. This was corrected and handled.
3. **Encoding Categorical Data**: After cleaning and imputation, I encoded categorical data for model training.
### Model Training
I trained the following models:
1. **RandomForestClassifier**:
```python
r_mod = RandomForestClassifier(n_estimators=100, random_state=42)
```
2. **XGBoost Classifier**:
```python
x_mod = XGBClassifier(objective='multi:softmax', n_estimators=100, random_state=42)
```
3. **Optimized XGBoost Classifier**:
```python
xm = XGBClassifier(objective='multi:softmax', n_estimators=500, learning_rate=0.5, random_state=42)
```
### Results
- **Random Forest Results**:
- Accuracy: 0.8399
- Precision: 0.8468
- Recall: 0.8399
- F1-Score: 0.8252
- **XGBoost Results**:
- Accuracy: 0.8910
- Precision: 0.8899
- Recall: 0.8910
- F1-Score: 0.8865
### Submission Files
The results were saved in the following files:
- `submission3.csv`
- `submission4.csv`
- `submission5.csv`
## How to Run
1. Clone this repository.
2. Open the `living-type-classification-from-codon-usage.ipynb` notebook.
3. Load the cleaned datasets: `train_replaced_missing_values.csv` and `dna_test - dna_test.csv.csv`.
4. Replace the train data path under the "Understanding Data" section and the test data path under the "Test Data Preparation" section of the notebook (use the `read_csv()` function).
5. Run the notebook.
Alternatively, you can start from scratch by using the raw data files (`dna_train.csv` and `dna_test.csv`) and following the cleaning and preprocessing steps in the notebook.