如何按N天分段计算并可视化线性回归曲线？

Got it, let's walk through exactly how to implement this segmented linear regression and visualize the results. You already have a solid slope/intercept calculator, so we'll build on that with a loop to handle the N-day chunks, then plot everything together.

Step 1: Define Your Segmentation Logic #

First, pick your segment size N (e.g., 7 for weekly segments). We'll split your time series into consecutive chunks of N days, and automatically handle the final partial chunk if your total data length isn't a perfect multiple of N.

Step 2: Full Implementation Code #

Here's how to wrap your existing function in a loop to compute each segment's regression line, then store all results for plotting:

import numpy as np
import pandas as pd
import matplotlib.pyplot as plt

# Your existing slope/intercept function (unchanged)
def slope_intercept(x_val, y_val):
    x = np.array(x_val)
    y = np.array(y_val)
    m = ( ( (np.mean(x)*np.mean(y) ) - np.mean(x*y)) / ( ( np.mean(x)*np.mean(x)) - np.mean(x*x)))
    m = round(m,2)
    b=(np.mean(y)-np.mean(x)*m)
    b=round(b,2)
    return m,b

# Example setup (replace with your actual DataFrame)
# Assume `future` is your time series DataFrame with 'close' column and sequential index
# future = pd.read_csv('your_data.csv', index_col='date', parse_dates=True)

# Set your desired segment size
N = 5  # Adjust this to your preferred number of days per segment

# Initialize a list to store each segment's regression data
regression_segments = []

# Iterate over each N-day chunk
total_days = len(future)
for i in range(0, total_days, N):
    # Get bounds for the current segment (handle final partial chunk)
    end_idx = min(i + N, total_days)
    segment = future.iloc[i:end_idx]
    
    # Extract x (index values) and y (close prices) for the segment
    x_vals = segment.index.tolist()
    y_vals = segment['close'].tolist()
    
    # Calculate slope and intercept using your function
    m, b = slope_intercept(x_vals, y_vals)
    
    # Generate regression line points for this segment
    segment['reg_line'] = [m * x + b for x in x_vals]
    
    # Add the segment to our list
    regression_segments.append(segment)

# Combine segments into a single DataFrame (optional but handy for analysis)
combined_reg_data = pd.concat(regression_segments)

Step 3: Visualize the Results #

Now let's plot the original close prices alongside all segmented regression lines. We'll loop through each segment to plot its line—you can even color-code segments if you want to distinguish them clearly:

plt.figure(figsize=(12,6))

# Plot original close price data
plt.plot(future.index, future['close'], label='Original Close Price', color='blue', alpha=0.5)

# Plot each segmented regression line
for seg_num, segment in enumerate(regression_segments):
    plt.plot(segment.index, segment['reg_line'], label=f'Segment {seg_num+1} Regression', linestyle='--')

plt.title(f'Segmented Linear Regression (Every {N} Days)')
plt.xlabel('Date/Index')
plt.ylabel('Close Price')
plt.legend()
plt.grid(True)
plt.show()

Notes for Datetime Indices #

If your future DataFrame uses datetime indices instead of integers, convert dates to numeric ordinals for the regression calculation (since linear regression requires numeric x-values):

# Replace x_vals with this for datetime indices
x_vals = segment.index.map(pd.Timestamp.toordinal).tolist()
# The plot will still use the original datetime index for readability

Edge Case Handling #

• Partial final segment: Our code uses min(i + N, total_days) to ensure we don't go out of bounds, so even the last chunk with fewer than N days gets a regression line.
• Skip partial segments: If you don't want to include partial chunks, add a check inside the loop: if end_idx - i < N: break

内容的提问来源于stack exchange，提问作者hkml