Long-Term-Care-Aggregated-Data.py

# -*- coding: utf-8 -*-
"""Long-Term-Care-Aggregated-Data.ipynb

Automatically generated by Colaboratory.

Original file is located at
    https://colab.research.google.com/drive/14YdgB8b4TtNetbHpTstGH7W4DLa3Yyxq
"""

from datasets import GeneratorBasedBuilder, DownloadManager, DatasetInfo, Array3D, BuilderConfig, SplitGenerator, Version
from datasets.features import Features, Value, Sequence
import datasets
import pandas as pd
import json
import zipfile
from PIL import Image
import numpy as np
import io
import csv
import json
import os
from typing import List
import datasets
import logging

_CITATION = """\
@misc{long_term_care_aggregated_dataset,
title = {Long-Term Care Aggregated Dataset},
author = {Kao, Hsuan-Chen (Justin)},
year = {2024},
publisher = {Hugging Face},
url = {https://github.com/justinkao44/STA663_Project_1},
}
"""

_DESCRIPTION = """\
The Long-Term Care Aggregated Dataset is a collection of insurance data for 'incidence' and 'termination' categories.
It is compiled from Long Term Care insurance policies data, providing insights into trends and patterns in insurance claims
and terminations. This dataset can be used for actuarial analysis, risk assessment, and to inform insurance product development.
"""

_HOMEPAGE = "https://github.com/justinkao44/STA663_Project_1"

_LICENSE = "Apache-2.0"

_URLS = {
    "train_incidence": "https://raw.githubusercontent.com/justinkao44/STA663_Project_1/main/train_filtered_incidence_df.csv",
    "train_termination": "https://raw.githubusercontent.com/justinkao44/STA663_Project_1/main/train_filtered_termination_df.csv",
    "validation_incidence": "https://raw.githubusercontent.com/justinkao44/STA663_Project_1/main/validation_filtered_incidence_df.csv",
    "validation_termination": "https://raw.githubusercontent.com/justinkao44/STA663_Project_1/main/validation_filtered_termination_df.csv",
}


class LongTermCareAggregatedData(datasets.GeneratorBasedBuilder):
    """Dataset for insurance 'incidence' and 'termination' data."""

    BUILDER_CONFIGS = [
        datasets.BuilderConfig(name="incidence", version=datasets.Version("1.0.0"), description="This part of the dataset includes incidence features"),
        datasets.BuilderConfig(name="termination", version=datasets.Version("1.0.0"), description="This part of the dataset includes termination features"),
    ]

    def _info(self):
        if self.config.name == "incidence":
            features = datasets.Features({
                "Group_Indicator": datasets.Value("string"),
                "Gender": datasets.Value("string"),
                "Issue_Age_Bucket": datasets.Value("string"),
                "Incurred_Age_Bucket": datasets.Value("string"),
                "Issue_Year_Bucket": datasets.Value("string"),
                "Policy_Year": datasets.Value("string"),
                "Marital_Status": datasets.Value("string"),
                "Premium_Class": datasets.Value("string"),
                "Underwriting_Type": datasets.Value("string"),
                "Coverage_Type_Bucket": datasets.Value("string"),
                "Tax_Qualification_Status": datasets.Value("string"),
                "Inflation_Rider": datasets.Value("string"),
                "Rate_Increase_Flag": datasets.Value("string"),
                "Restoration_of_Benefits": datasets.Value("string"),
                "NH_Orig_Daily_Ben_Bucket": datasets.Value("string"),
                "ALF_Orig_Daily_Ben_Bucket": datasets.Value("string"),
                "HHC_Orig_Daily_Ben_Bucket": datasets.Value("string"),
                "NH_Ben_Period_Bucket": datasets.Value("string"),
                "ALF_Ben_Period_Bucket": datasets.Value("string"),
                "HHC_Ben_Period_Bucket": datasets.Value("string"),
                "NH_EP_Bucket": datasets.Value("string"),
                "ALF_EP_Bucket": datasets.Value("string"),
                "HHC_EP_Bucket": datasets.Value("string"),
                "Region": datasets.Value("string"),
                "Active_Exposure": datasets.Value("float64"),
                "Total_Exposure": datasets.Value("float64"),
                "Claim_Count": datasets.Value("int32"),
                "Count_NH": datasets.Value("int32"),
                "Count_ALF": datasets.Value("int32"),
                "Count_HHC": datasets.Value("int32"),
                "Count_Unk": datasets.Value("int32"),
            })
        elif self.config.name == "termination":
            features = datasets.Features({
                "Gender": datasets.Value("string"),
                "Incurred_Age_Bucket": datasets.Value("string"),
                "Incurred_Year_Bucket": datasets.Value("string"),
                "Claim_Type": datasets.Value("string"),
                "Region": datasets.Value("string"),
                "Diagnosis_Category": datasets.Value("string"),
                "Claim_Duration": datasets.Value("int64"),
                "Exposure": datasets.Value("int64"),
                "Deaths": datasets.Value("int64"),
                "Recovery": datasets.Value("int64"),
                "Terminations": datasets.Value("int64"),
                "Benefit_Expiry": datasets.Value("int64"),
                "Others_Terminations": datasets.Value("int64"),
            })
        else:
            raise ValueError(f"BuilderConfig name not recognized: {self.config.name}")

        return datasets.DatasetInfo(
            description=_DESCRIPTION,
            features=features,
            supervised_keys=None,
            homepage="https://www.soa.org/resources/experience-studies/2020/2000-2016-ltc-aggregate-database/",
            citation="Please cite this dataset as: Society of Actuaries (SOA). (2020). Long Term Care Insurance Aggregate Experience Data, 2000-2016."
        )


    def _split_generators(self, dl_manager):
        downloaded_files = dl_manager.download(_URLS)

        # 基于配置名确定文件路径
        train_file = downloaded_files[f"train_{self.config.name}"]
        validation_file = downloaded_files[f"validation_{self.config.name}"]

        return [
            datasets.SplitGenerator(
                name=datasets.Split.TRAIN,
                gen_kwargs={
                    "data_file": train_file,
                    "split": self.config.name,  # 使用配置名作为 split 参数（如果需要）
                },
            ),
            datasets.SplitGenerator(
                name=datasets.Split.VALIDATION,
                gen_kwargs={
                    "data_file": validation_file,
                    "split": self.config.name,  # 使用配置名作为 split 参数（如果需要）
                },
            ),
        ]

    def _generate_examples(self, data_file, split):
        # Read the CSV file for the given configuration
        dataframe = pd.read_csv(data_file)

        # Determine the feature columns based on the configuration
        feature_columns = self._get_feature_columns_by_config()

        # Yield examples using the determined feature columns
        for idx, row in dataframe.iterrows():
            feature_dict = {column: row[column] for column in feature_columns if column in dataframe.columns}
            yield idx, feature_dict

    def _get_feature_columns_by_config(self):
        if self.config.name == "incidence":
            return [
                "Group_Indicator", "Gender", "Issue_Age_Bucket", "Incurred_Age_Bucket",
                "Issue_Year_Bucket", "Policy_Year", "Marital_Status", "Premium_Class",
                "Underwriting_Type", "Coverage_Type_Bucket", "Tax_Qualification_Status",
                "Inflation_Rider", "Rate_Increase_Flag", "Restoration_of_Benefits",
                "NH_Orig_Daily_Ben_Bucket", "ALF_Orig_Daily_Ben_Bucket", "HHC_Orig_Daily_Ben_Bucket",
                "NH_Ben_Period_Bucket", "ALF_Ben_Period_Bucket", "HHC_Ben_Period_Bucket",
                "NH_EP_Bucket", "ALF_EP_Bucket", "HHC_EP_Bucket", "Region",
                "Active_Exposure", "Total_Exposure", "Claim_Count", "Count_NH", "Count_ALF", "Count_HHC", "Count_Unk",
            ]
        elif self.config.name == "termination":
            return [
                "Gender", "Incurred_Age_Bucket", "Incurred_Year_Bucket", "Claim_Type",
                "Region", "Diagnosis_Category", "Claim_Duration", "Exposure", "Deaths",
                "Recovery", "Terminations", "Benefit_Expiry", "Others_Terminations",
            ]
        else:
            raise ValueError(f"Config name not recognized: {self.config.name}")