Model.py

Introduction

The model.py file (opens in a new tab) in basic-coin-prediction-node consists of several key components:

Imports and Configuration: Sets up necessary libraries and configuration variables.
Paths Configuration: Generates paths for storing data dynamically based on coin symbols.
Downloading Data: Downloads historical price data for the specified symbols, intervals, years, and months.
Formatting Data: Reads, formats, and saves the downloaded data as CSV files.
Training the Model: Trains a linear regression model on the formatted price data and saves the trained model.

While the import and path configuration processes are straightforward, downloading and formatting the data, as well as training the model, require specific steps.

This documentation will guide you through creating models for different coins, making it easy to extend the script for general-purpose use.

Downloading the Data

The download_data (opens in a new tab) function is designed to automate the process of downloading historical market data from Binance, a popular cryptocurrency exchange. This function focuses on fetching data for a specified set of symbols (in this case, the trading pair "ETHUSDT") across various time intervals and storing them in a defined directory.

How to Use for Downloading Data of Any Coin

Update the Symbols List

Replace ["ETHUSDT"] with the desired trading pair(s), e.g., ["BTCUSDT", "LTCUSDT"].

Adjust Time Intervals

Modify the intervals list if you need different time intervals. Binance supports various intervals like ["1m", "5m", "1h", "1d", "1w", "1M"].

Extend Date Ranges

Update the years and months lists to match the historical range you need.

Define the Download Path

Ensure binance_data_path is set to the directory where you want the data to be saved.

Here’s a quick example of how to adjust the script for downloading data for multiple trading pairs:

def download_data():
    cm_or_um = "um"
    symbols = ["BTCUSDT", "LTCUSDT"]  # Updated symbols
    intervals = ["1d"]
    years = ["2020", "2021", "2022", "2023", "2024"]
    months = ["01", "02", "03", "04", "05", "06", "07", "08", "09", "10", "11", "12"]
    download_path = binance_data_path
    download_binance_monthly_data(
        cm_or_um, symbols, intervals, years, months, download_path
    )
    print(f"Downloaded monthly data to {download_path}.")
    current_datetime = datetime.now()
    current_year = current_datetime.year
    current_month = current_datetime.month
    download_binance_daily_data(
        cm_or_um, symbols, intervals, current_year, current_month, download_path
    )
    print(f"Downloaded daily data to {download_path}.")

Formatting the Data

The format_data (opens in a new tab) function processes raw data files downloaded from Binance, transforming them into a consistent format for analysis. Here are the key steps:

File Handling:
- Lists and sorts all files in the binance_data_path directory.
- Exits if no files are found.
Initialize DataFrame:
- An empty DataFrame price_df is created to store the combined data.
Process Each File:
- Filters for .zip files and reads the contained CSV file.
- Retains the first 11 columns and renames them to: ["start_time", "open", "high", "low", "close", "volume", "end_time", "volume_usd", "n_trades", "taker_volume", "taker_volume_usd"].
- Sets the DataFrame index to the end_time column, converted to a timestamp.
Concatenate Data:
- Combines data from each file into the price_df DataFrame.
Sort and Save:
- Sorts the final DataFrame by date and saves it to training_price_data_path.

Column Descriptions

start_time: The start of the trading period.
open: Opening price.
high: Highest price during the period.
low: Lowest price during the period.
close: Closing price.
volume: Trading volume.
end_time: End of the trading period.
volume_usd: Trading volume in USD.
n_trades: Number of trades.
taker_volume: Taker buy volume.
taker_volume_usd: Taker buy volume in USD.

This function consolidates and formats the historical price data, making it ready for analysis or machine learning tasks.

Training the Model

The train_model (opens in a new tab) function trains a linear regression model using historical price data and saves the trained model to a file. Here's a breakdown of the process:

Load the Data:
- Reads the price data from a CSV file specified by training_price_data_path.
Prepare the DataFrame:
- Converts the date column to a timestamp and stores it as a numerical value.
- Computes the average price using the open, close, high, and low columns.
Reshape Data for Regression:
- Extracts the date column as the feature (x) and the computed average price as the target (y).
- Reshapes these arrays to the format expected by scikit-learn.
Split the Data:
- Splits the data into a training set and a test set using an 80/20 split. However, the test set is not used further in this function.
Train the Model:
- Initializes and trains a LinearRegression model using the training data.
Save the Model:
- Creates the directory for the model file if it doesn't exist.
- Saves the trained model to a file specified by model_file_path using pickle.
Print Confirmation:
- Prints a message indicating that the trained model has been saved.

Modifying the Function for Different Models

💡

Understanding the Target Variable in Regression Models

The target variable (y) in regression models is a critical component that determines the type of analysis and predictions that can be performed. In this context, the target variable represents continuous data—in this case, the average of financial metrics such as the open, close, high, and low prices over time. The y-axis points on the regression graph correspond to these continuous values, which can take on any numerical value within a defined range.

To change the model used for training, replace the LinearRegression model with another machine learning algorithm. Here is an example:

Using Polynomial Regression

from sklearn.preprocessing import PolynomialFeatures
from sklearn.linear_model import LinearRegression
from sklearn.pipeline import make_pipeline
import os
import pickle
import pandas as pd
from sklearn.model_selection import train_test_split
 
def train_model():
    # Load the eth price data
    price_data = pd.read_csv(training_price_data_path)
    df = pd.DataFrame()
 
    # Convert 'date' to a numerical value (timestamp) we can use for regression
    df["date"] = pd.to_datetime(price_data["date"])
    df["date"] = df["date"].map(pd.Timestamp.timestamp)
 
    # Calculate the mean price as the target variable
    df["price"] = price_data[["open", "close", "high", "low"]].mean(axis=1)
 
    # Reshape the data to the shape expected by sklearn
    x = df["date"].values.reshape(-1, 1)
    y = df["price"].values.reshape(-1, 1)
 
    # Split the data into training set and test set
    x_train, _, y_train, _ = train_test_split(x, y, test_size=0.2, random_state=0)
 
    # Create a pipeline that first transforms the input data to polynomial features and then fits a linear model
    model = make_pipeline(PolynomialFeatures(degree=2), LinearRegression())
    model.fit(x_train, y_train)
 
    # Create the model's parent directory if it doesn't exist
    os.makedirs(os.path.dirname(model_file_path), exist_ok=True)
 
    # Save the trained model to a file
    with open(model_file_path, "wb") as f:
        pickle.dump(model, f)
 
    print(f"Trained polynomial regression model saved to {model_file_path}")

Price Prediction Worker Reputers