如何将Python循环输出的分布拟合结果转换为DataFrame格式？

Here's how you can tweak your existing code to gather all results into a structured DataFrame instead of printing them line by line:

Modified Code #

import pandas as pd

def distribution_selection(csv_file=None, product_column=None, demand=None):
    # Read the CSV into a DataFrame (fixes the undefined df_all reference in your original code)
    df_all = pd.read_csv(csv_file) if csv_file else pd.DataFrame()
    
    results = []  # Store each product's data as a dictionary here
    
    for num in df_all[product_column]:
        # ... (keep your existing logic to compute best_fit, likelihoods, and parameters like mean, std, p_nbinom, r_binom, lambda_)
        
        # Start with common fields for every product
        product_result = {
            "Products": num,
            "BestFit": best_fit,
            "Likelihood": likelihoods[best_fit]
        }
        
        # Add parameters based on the best-fit distribution
        if best_fit == "norm":
            product_result["ParameterA"] = mean
            product_result["ParameterB"] = std
        elif best_fit == "nbinom":
            product_result["ParameterA"] = p_nbinom
            product_result["ParameterB"] = r_binom
        elif best_fit == "poisson":
            product_result["ParameterA"] = lambda_
            product_result["ParameterB"] = pd.NA  # Mark missing second parameter
        
        results.append(product_result)
    
    # Convert the list of dictionaries to a DataFrame
    output_df = pd.DataFrame(results)
    
    # Reorder columns to match your desired structure
    output_df = output_df[["Products", "BestFit", "Likelihood", "ParameterA", "ParameterB"]]
    
    # Print the final DataFrame (or return it for further use)
    print(output_df)
    return output_df

Key Changes Breakdown #

• Result Collection: Instead of printing immediately, we store each product's data in a dictionary and add it to a list. This makes converting to a DataFrame straightforward.
• Parameter Handling: For each distribution type, we map the computed parameters to ParameterA and ParameterB. For Poisson (which only has one parameter), we use pd.NA to represent the missing value, which shows up as NA in the final output.
• CSV Reading: Added a line to load the CSV file into df_all since your original code referenced this variable but didn't define it in the function.
• Column Ordering: Reordered the DataFrame columns to exactly match the structure you requested.

Example Output #

When you run the function, you'll get a neatly formatted DataFrame like this:

内容的提问来源于stack exchange，提问作者Salih