求助：基于振荡值为DataFrame标记状态列的实现方法

Alright, let's break down how to solve this problem step by step. From your description, we need to mark alternating positive/negative cycles in a DataFrame, where each cycle starts when column Y transitions from 0 to a positive value, ends at the next 0 (right before it would switch to negative), and repeats this pattern for negative cycles afterward.

Step 1: Prepare Sample Data #

First, let's create a sample DataFrame to demonstrate the logic. This mimics the oscillating pattern you described:

import pandas as pd
import numpy as np

# Set seed for reproducibility
np.random.seed(42)

# Generate X values and oscillating Y values (0 → positive → 0 → negative → 0...)
x = np.linspace(0, 20, 100)
y = np.where(
    (x >= 0) & (x < 5), np.random.uniform(0.5, 2, 25),
    np.where(
        (x >= 5) & (x < 10), 0,
        np.where(
            (x >= 10) & (x < 15), np.random.uniform(-2, -0.5, 25),
            0
        )
    )
)

df = pd.DataFrame({'X': x, 'Y': y})

Step 2: Identify Cycle Boundaries #

We need to detect when Y transitions between 0 and non-zero values, which marks the start/end of each cycle. Here's how to do it:

# 1. Label Y's current state (positive, negative, zero)
df['y_state'] = np.where(df['Y'] > 0, 'pos', 
                         np.where(df['Y'] < 0, 'neg', 'zero'))

# 2. Find indices where the state changes (e.g., zero → pos, pos → zero)
state_changes = df['y_state'] != df['y_state'].shift()
change_points = df[state_changes].index.tolist()

Step 3: Define and Label Each Cycle #

Now we'll iterate through the state change points to define each cycle's start, end, and type (positive/negative), then apply these labels to the DataFrame:

# Initialize list to store cycle details and label column
cycles = []
df['cycle_label'] = 'none'

# Iterate through state changes to map cycles
for i in range(len(change_points) - 1):
    start_idx = change_points[i]
    end_idx = change_points[i+1]
    start_state = df.loc[start_idx, 'y_state']
    end_state = df.loc[end_idx, 'y_state']
    
    # Check if this is a valid positive cycle (zero → pos → zero)
    if start_state == 'zero' and df.loc[start_idx+1, 'y_state'] == 'pos' and end_state == 'zero':
        cycle_name = f"positive_cycle_{len(cycles)+1}"
        df.loc[start_idx:end_idx, 'cycle_label'] = cycle_name
        cycles.append({'type': 'positive', 'start': start_idx, 'end': end_idx})
    # Check if this is a valid negative cycle (zero → neg → zero)
    elif start_state == 'zero' and df.loc[start_idx+1, 'y_state'] == 'neg' and end_state == 'zero':
        cycle_name = f"negative_cycle_{len(cycles)+1}"
        df.loc[start_idx:end_idx, 'cycle_label'] = cycle_name
        cycles.append({'type': 'negative', 'start': start_idx, 'end': end_idx})

Step 4: Verify the Result #

You can check the labeled DataFrame with:

# Print first 30 rows (covers first positive cycle and zero transition)
print(df[['X', 'Y', 'cycle_label']].head(30))

# Print last 30 rows (covers negative cycle and final zero)
print(df[['X', 'Y', 'cycle_label']].iloc[70:100])

Edge Case Handling #

• If your dataset starts with a non-zero Y value (not a zero → positive transition), the first partial cycle will be labeled as 'none' (you can adjust this if needed to mark it as incomplete).
• If the dataset ends with a non-zero Y value (no closing zero), that cycle will remain unlabeled; you can extend the logic to mark it as incomplete_{type}_cycle if required.

内容的提问来源于stack exchange，提问作者Life2Day