data_helper.py

import os
import glob

import numpy as np
import pandas as pd
from datetime import datetime, timedelta


class DataHelper:
    def __init__(self, symbol, timeframe, data_dir="data/btcusd"):
        self.symbol = symbol
        self.timeframe = timeframe
        self.data_dir = data_dir

    def fetch_historical_data(self):
        file_pattern = os.path.join(self.data_dir, f"{self.symbol}-{self.timeframe}-*.csv")
        csv_files = glob.glob(file_pattern)
        sorted_csv_files = sorted(csv_files)

        df_list = []
        for csv_file in sorted_csv_files:
            try:
                df = pd.read_csv(csv_file)
                df['timestamp'] = pd.to_datetime(df['timestamp'], unit='ms')
                df_list.append(df)
            except Exception as e:
                print(f"Error processing file {csv_file}: {e}")

        return pd.concat(df_list, ignore_index=True)

    def get_halving_date(self, year):
        halving_dates = self.halving_dates()
        for date in halving_dates:
            if date.year == year:
                return date
        raise ValueError(f"No halving date found for the year {year}")

    def halving_dates(self):
        return [datetime(2012, 11, 28), datetime(2016, 7, 9), datetime(2020, 5, 11), datetime(2024, 5, 12),
                datetime(2028, 5, 12)]

    def days_since_last_halving(self, current_date):
        halving_dates = self.halving_dates()
        past_halvings = [date for date in halving_dates if date < current_date]
        if not past_halvings:
            return 0
        last_halving = max(past_halvings)
        return (current_date - last_halving).days

    def generate_future_features(self, data, features, days=90):
        # Get the last known value for each feature
        last_values = data[features].iloc[-1]

        # Calculate historical percentage changes
        pct_changes = data['close'].pct_change().dropna()

        # Focus on percentage changes around all-time highs (e.g., within 5% of the max price)
        all_time_high = data['close'].max()
        threshold = all_time_high * 0.05
        at_high_pct_changes = pct_changes[data['close'] > (all_time_high - threshold)]

        # Calculate the average percentage change around all-time highs
        avg_pct_change_at_high = at_high_pct_changes.mean()

        # Create a DataFrame to hold future feature values
        future_data = pd.DataFrame(index=pd.date_range(start=data['timestamp'].iloc[-1] + timedelta(days=1), periods=days))

        # Generate future feature values using the average percentage change
        for feature in features:
            future_values = [last_values[feature]]
            for _ in range(1, days):
                future_values.append(future_values[-1] * (1 + avg_pct_change_at_high))
            future_data[feature] = future_values

        return future_data

    def generate_features_to_halving(self, data, features):
        # Get the last known date and value for each feature
        last_known_date = data['timestamp'].iloc[-1]
        last_values = data[features].iloc[-1]

        # Determine the halving date
        halving_date = self.get_halving_date(2024)  # Replace with actual function to get halving date

        # Calculate the number of days to the halving date
        days_to_halving = (halving_date - last_known_date).days

        # Calculate historical percentage changes
        pct_changes = data['close'].pct_change().dropna()

        # Focus on percentage changes around all-time highs (for example)
        all_time_high = data['close'].max()
        threshold = all_time_high * 0.05
        at_high_pct_changes = pct_changes[data['close'] > (all_time_high - threshold)]

        # Calculate the average percentage change around all-time highs
        avg_pct_change_at_high = at_high_pct_changes.mean()

        # Create a DataFrame to hold future feature values
        future_data = pd.DataFrame(index=pd.date_range(start=last_known_date + timedelta(days=1), periods=days_to_halving))

        # Generate future feature values using the average percentage change
        for feature in features:
            future_values = [last_values[feature]]
            for _ in range(1, days_to_halving):
                future_values.append(future_values[-1] * (1 + avg_pct_change_at_high))
            future_data[feature] = future_values

        return future_data