import os

import pandas as pd

import numpy as np

import PyPDF2

import tensorflow as tf

from tensorflow.keras.models import Sequential

from tensorflow.keras.layers import LSTM, Dense, Dropout

from tensorflow.keras.preprocessing.sequence import TimeseriesGenerator

# PDF에서 데이터 추출 함수

def extract_text_from_pdfs(pdf_folder):

    data = []

    for file in os.listdir(pdf_folder):

        if file.endswith(".pdf"):

            with open(os.path.join(pdf_folder, file), "rb") as pdf_file:

                reader = PyPDF2.PdfReader(pdf_file)

                text = " ".join([page.extract_text() for page in reader.pages if page.extract_text()])

                data.append(text)

    return data

# 데이터 전처리 함수

def preprocess_data(text_data):

    processed_data = []

    for text in text_data:

        lines = text.split("\n")

        for line in lines:

            values = line.split()

            if len(values) > 1:

                processed_data.append(values)

    return pd.DataFrame(processed_data)

# LSTM 모델 생성 함수

def create_lstm_model(input_shape):

    model = Sequential([

        LSTM(50, activation='relu', return_sequences=True, input_shape=input_shape),

        Dropout(0.2),

        LSTM(50, activation='relu'),

        Dropout(0.2),

        Dense(1)

    ])

    model.compile(optimizer='adam', loss='mse')

    return model

# 예측 실행 함수

def train_and_predict_lstm(data, lookback=5):

    data = np.array(data, dtype=np.float32)

    data = data.reshape(-1, 1)

    generator = TimeseriesGenerator(data, data, length=lookback, batch_size=1)

    model = create_lstm_model((lookback, 1))

    model.fit(generator, epochs=20, verbose=1)

    predictions = model.predict(data[-lookback:].reshape(1, lookback, 1))

    return predictions

# 실행 예시

pdf_folder = "./pdf_data"  # PDF 파일이 저장된 폴더 경로

pdf_texts = extract_text_from_pdfs(pdf_folder)

df = preprocess_data(pdf_texts)

df.to_csv("processed_data.csv", index=False)

# 데이터 가공 및 예측 (예시로 1열의 데이터를 사용)

if df.shape[1] > 1:

    target_column = df.iloc[:, 1].dropna().astype(float)

    prediction = train_and_predict_lstm(target_column)

    print("예측 결과:", prediction)