Cargo Loading Volume Prediction Pseudo Code

Pseudo Code for Ship Cargo Loading Volume Prediction

This pseudo code translates the provided Python script into a structured format, outlining the major steps taken to preprocess the data and generate features for a machine learning task focused on predicting cargo loading volumes.

Imports and Warnings Management

1. Import Required Libraries:
   • Import data manipulation libraries: Pandas, NumPy
   • Import machine learning libraries: LightGBM, XGBoost
   • Import metrics and validation techniques: Mean Squared Error, StratifiedKFold, KFold, GroupKFold
   • Manage warnings by ignoring them.

Constants

• Define a constant for the label (target variable): LABEL is set to '装载量'.

Data Loading

1. Read Training and Testing Data:

   • Load the training dataset from a specified path with appropriate encoding.
   • Load the testing dataset from a specified path with appropriate encoding.

2. Combine Training and Testing Data:

   • Concatenate training and testing DataFrames into one.

Data Preprocessing

1. Handle Missing Values and Type Conversion:

   • Replace ' None' in departure time with NaN and convert to float.
   • Calculate the time difference: time_diff as the difference between departure and arrival times.

2. Convert Timestamps:

   • Convert '进泊时间' and '离泊时间' to datetime format using Unix timestamps.

Feature Engineering

1. Extract Time Features:

   • For each time variable ('进泊时间', '离泊时间'):
     • Extract hour, day of week, and quarter.

2. Encode Categorical Variables:

   • Factorize categorical codes (A and B) to create numeric representations.

3. Generate Statistical Features:

   • For selected numeric features '载重吨' and 'time_diff':
     • For each categorical feature ('A_le', 'B_le', 'AB_le'):
       • Calculate and store mean, standard deviation, max, and min.

4. Additional Statistical Relationships:

   • For each numeric feature, calculate mean, standard deviation, max, and min against each categorical feature.

5. Location Feature Extraction:

   • Parse location strings into separate float values (loc0 and loc1).

6. Combination Features:

   • Calculate area as the product of ship length and width.
   • Count occurrences grouped by ship ID and ship type codes.

7. Group Counts:

   • For each unique value of ship ID, width, and loaded tonnage, count occurrences.

8. Longitude/Latitude Ratio:

   • Calculate a derived geographic feature as the ratio of loc0 to loc1.

Pseudo Code Representation

BEGIN

    # Import libraries
    IMPORT pandas AS pd
    IMPORT numpy AS np
    IMPORT lightgbm AS lgb
    IMPORT xgboost AS xgb
    IMPORT mean_squared_error AS mse
    IMPORT StratifiedKFold, KFold, GroupKFold

    # Suppress warnings
    ENABLE warnings
    SET warnings TO ignore

    # Define LABEL
    SET LABEL = '装载量'

    # Load datasets
    df_train = READ_CSV('./data/船舶装卸货量预测-训练集-20240611.csv', encoding='gbk')
    df_test = READ_CSV('./data/船舶装卸货量预测-测试集X-20240611.csv', encoding='gbk')

    # Combine datasets
    df = CONCATENATE(df_train, df_test)

    # Preprocess data
    REPLACE ' None' WITH NaN IN df['离泊时间']
    CONVERT df['离泊时间'] TO float
    df['time_diff'] = df['离泊时间'] - df['进泊时间']

    CONVERT df['进泊时间'] AND df['离泊时间'] TO datetime

    # Extract time features
    FOR each f IN ['进泊时间', '离泊时间']:
        df[f + '_hour'] = EXTRACT HOUR FROM df[f]
        df[f + '_dayofweek'] = EXTRACT DAY OF WEEK FROM df[f]
        df[f + '_quarter'] = EXTRACT QUARTER FROM df[f]

    # Non-numeric encoding
    df['A_le'] = FACTORIZE(df['船舶类型代码A'])
    df['B_le'] = FACTORIZE(df['船舶类型代码B'])

    # Generate cross statistics
    FOR each num_f IN ['载重吨', 'time_diff']:
        FOR each cat_f IN ['A_le', 'B_le', 'AB_le']:
            df[cat_f + '_' + num_f + '_mean'] = GROUPBY(cat_f, df[num_f], APPLY MEAN)
            # Repeat for std, max, min using a similar structure

    # Create location features
    df['loc0'] = MAP(df['泊位位置'], EXTRACTION FUNCTION)
    df['loc1'] = MAP(df['泊位位置'], EXTRACTION FUNCTION)

    # Combination features
    df['面积'] = df['船长'] * df['船宽']
    df['船舶ID_航次ID_count'] = GROUPBY(df['船舶ID'], COUNT(df['航次ID']))
    
    # Repeat for other combination statistics using GROUPBY

    # Calculate longitude/latitude
    df['longitude/Latitude'] = df['loc0'] / df['loc1']

END

Notes

• Each segment of the code has been translated into its logical progression.
• The pseudo code is designed to present the process in an understandable manner, facilitating discussions or implementations without direct reference to programming specifics.