app.py

import streamlit as st
import numpy as np
import tensorflow as tf
import joblib
from PIL import Image

from tensorflow.keras.models import load_model

st.set_page_config(page_title="Nifty50 Predictor", layout="centered")
st.title("📈 PricePulse")
st.subheader("A Nifty50 Price Predictor", divider='grey')
image = Image.open('nifty50stock.jpg')
st.image(image)

# Load model and scaler
@st.cache_resource
def load_artifacts():
    model = load_model("final_nifty50_lstm.keras")
    scaler = joblib.load("scaler.pkl")
    return model, scaler

model, scaler = load_artifacts()

st.markdown("Enter the **last 3 closing prices** of NIFTY50 to predict the next price.")

col1, col2, col3 = st.columns(3)
with col1:
    price1 = st.number_input("Previous Day 3's Closing Price", value=22000.0, format="%.2f")
with col2:
    price2 = st.number_input("Previous Day 2's Closing Price", value=22100.0, format="%.2f")
with col3:
    price3 = st.number_input("Previous Day 1's Closing Price", value=22200.0, format="%.2f")

last3 = np.array([price1, price2, price3]).reshape(-1, 1)
last3_scaled = scaler.transform(last3).reshape(1, 3, 1)

# Predict next step
if st.button("🔮 Predict Next Day Price"):
    next_scaled = model.predict(last3_scaled, verbose=0)[0][0]
    next_price = scaler.inverse_transform([[next_scaled]])[0][0]
    st.success(f"Predicted Next Closing Price: ₹{next_price:.2f}")

# Predict next 10 steps recursively
if st.button("📈 Predict Next 10 Days"):
    seq = last3_scaled.copy()
    predictions = []
    for _ in range(10):
        pred = model.predict(seq, verbose=0)[0][0]
        predictions.append(pred)
        seq = np.append(seq[:, 1:, :], [[[pred]]], axis=1)
    inv_preds = scaler.inverse_transform(np.array(predictions).reshape(-1, 1)).flatten()
    st.subheader("Next 10 Predicted Prices")
    st.table({"Day": [f"Day {i+1}" for i in range(10)], "Predicted Price (₹)": inv_preds})

st.caption("Made with ❤️ by Sourish")