app.py

import pandas as pd
import gradio as gr
from sklearn.pipeline import Pipeline
from sklearn.preprocessing import StandardScaler
from sklearn.linear_model import LinearRegression
from transformers import pipeline
import numpy as np
import matplotlib.pyplot as plt
import wikipedia
from transformers import BertTokenizer, AutoModelForSeq2SeqLM, pipeline
from arabert.preprocess import ArabertPreprocessor

# Load the Excel sheet
df = pd.read_excel('stock_data.xlsx')
 

# Prepare the data for the model
X = df[['final price today', 'change today']].values
y = df['future_price'].values

"""
 Create a pipeline with a scaler and a linear regression model 
 scaler is used to standrlize the data and make the model more accurate. 
 and the linear regression model is used to predict the future price.

"""

model_pipeline = Pipeline([
    ('scaler', StandardScaler()),
    ('regressor', LinearRegression())
])
 
# Train the pipeline
model_pipeline.fit(X, y)

# Initialize the summary generation pipeline
model_name="malmarjeh/mbert2mbert-arabic-text-summarization"
preprocessor = ArabertPreprocessor(model_name="")
tokenizer = BertTokenizer.from_pretrained(model_name)
model = AutoModelForSeq2SeqLM.from_pretrained(model_name)
pipeline = pipeline("text2text-generation",model=model,tokenizer=tokenizer)

# function for building the plot
def plot_stock_prices(stock_name, current_price, future_price):
    plt.figure(figsize=(10, 20))

    # Plot the bar chart
    plt.bar(['Final Price Today', 'Predicted Future Price'], [current_price, future_price], color=['blue', 'green'])

    # Add labels and title
    plt.xlabel('Price Type')
    plt.ylabel('Price')
    plt.title(f'Stock Prices for {stock_name}')

    # Annotate the bars with the actual values
    plt.text(0, current_price, f'{current_price:.2f}', ha='center', va='bottom')
    plt.text(1, future_price, f'{future_price:.2f}', ha='center', va='bottom')

    # Adjust y-axis limits if the difference is very small
    if abs(current_price - future_price) < 1:
        plt.ylim(min(current_price, future_price) - 1, max(current_price, future_price) + 1)

    # Save the plot
    plt.savefig('graph.png')
    plt.close()

# Function to get stock information and predict future price and make the summary
def get_stock_info(stock_name):
    # check if the name in the list
    stock_info = df[df['name'] == stock_name]
    if stock_info.empty:

        # Return a message and an empty graph
        plt.figure(figsize=(10, 6))
        plt.text(0.5, 0.5, 'Stock not found', horizontalalignment='center', verticalalignment='center', fontsize=12)
        plt.axis('off')
        plt.savefig('not_found.png')
        plt.close()
        return {"Error": "Stock not found"}, 'not_found.png', 'training_plot.png'

    else:
        # Set Wikipedia language to Arabic
        wikipedia.set_lang("ar")
        try:
          # searching with the stock name and note that the use of the words "شركة" and "السعودية" is to make sure the search is about companies in saudi market
            wiki_summary = wikipedia.summary("شركة "+ stock_name + " السعودية", sentences=5)
        except wikipedia.exceptions.DisambiguationError as e:
            wiki_summary = f"Multiple entries found for {stock_name}: {e.options}"
        except wikipedia.exceptions.PageError:
            wiki_summary = "Wikipedia page not found for this company."
        # Setting the data to make the prediction and summary
        info = stock_info.iloc[0]
        current_price = info['final price today']
        change_today = info['change today']
        # making the prediction
        future_price = model_pipeline.predict([[current_price, change_today]])[0]
        summary_prompt = f"سهم شركة {stock_name} سعره الحالي{current_price} اليوم مع تغيير قدره {change_today}. ويتوقع ان يصبح سعره {future_price} والجدير بالذكر ان " 
        text = preprocessor.preprocess(wiki_summary)
        # building the summary
        summary = pipeline(text,
            pad_token_id=tokenizer.eos_token_id,
            num_beams=3,
            repetition_penalty=3.0,
            max_length=300,
            length_penalty=1.0,
            no_repeat_ngram_size = 3)[0]['generated_text']

        # Plotting the graph
        graph = plot_stock_prices('Example Stock', current_price, future_price)

        
        return {
            "اخر قيمة لليوم": current_price,
            "اعلى قيمة لليوم": info.get('highest price today', 'N/A'),
            "اقل قيمة لليوم": info.get('lowest price today', 'N/A'),
            "التغير": change_today,
            "نسبة التغير": info.get('percentage of change today', 'N/A'),
            "الحجم": info.get('size', 'N/A'),
            "اخر تحديث (بالساعة)": info.get('last update time', 'N/A'),
            "السعر المتوقع": future_price,
            "التوقع والملخص": summary_prompt + summary,
            "مقال ويكيبيديا الكامل": wiki_summary
        }, "graph.png"

# Create the Gradio interface
iface = gr.Interface(
    fn=get_stock_info,
    inputs=gr.Textbox(label="اسم السهم"),
    outputs=[gr.JSON(), gr.Image()],
    title="توقع اسعار سوق الاسهم السعودي",
    description="الرجاء ادخال اسم الشركة لاظهار التوقع ومعلومات عامة عن الشركة"

)
 
# Launch the interface with debug=True for error handling
iface.launch(debug=True)