Hamza012bce21 commited on
Commit
733a79d
Β·
verified Β·
1 Parent(s): 07acec0

Update app.py

Browse files
Files changed (1) hide show
  1. app.py +45 -70
app.py CHANGED
@@ -1,107 +1,82 @@
1
  import streamlit as st
2
  import numpy as np
3
  import pandas as pd
4
- import requests
5
- from tensorflow.keras.models import load_model
6
- import tensorflow as tf
7
- from sklearn.preprocessing import MinMaxScaler
8
- import joblib
9
- from vaderSentiment.vaderSentiment import SentimentIntensityAnalyzer
10
  import matplotlib.pyplot as plt
11
  import os
12
 
13
- # ------------------------------
14
- # Load Model & Scaler
15
- # ------------------------------
16
  try:
17
- model = load_model("model.h5", custom_objects={"mse": tf.keras.metrics.MeanSquaredError()})
18
- st.success("Model loaded successfully!")
19
- except Exception as e:
20
- st.error(f"Error loading model: {e}")
 
 
 
 
 
 
 
21
 
22
- scaler = joblib.load("scaler.pkl") if os.path.exists("scaler.pkl") else MinMaxScaler()
 
23
 
24
  # ------------------------------
25
- # Fetch PSX Data (AlphaVantage or Fallback)
26
  # ------------------------------
27
- API_KEY = os.getenv("ALPHAVANTAGE_API_KEY", None)
28
-
29
- def get_psx_data(symbol="HBL"):
30
- if API_KEY:
31
- url = f"https://www.alphavantage.co/query?function=TIME_SERIES_DAILY&symbol={symbol}.PSX&apikey={API_KEY}"
32
- try:
33
- r = requests.get(url).json()
34
- data = r["Time Series (Daily)"]
35
- df = pd.DataFrame(data).T
36
- df.index = pd.to_datetime(df.index)
37
- df = df.sort_index()
38
- df = df[["4. close"]].rename(columns={"4. close": "Close"})
39
- return df
40
- except:
41
- pass
42
 
43
- # Fallback simulated data
 
 
 
44
  dates = pd.date_range(end=pd.Timestamp.today(), periods=200)
45
  prices = np.linspace(100, 150, 200) + np.random.normal(0, 2, 200)
46
  df = pd.DataFrame({"Close": prices}, index=dates)
47
  return df
48
 
49
  # ------------------------------
50
- # News Sentiment
51
- # ------------------------------
52
- NEWS_KEY = os.getenv("NEWSAPI_KEY", None)
53
- analyzer = SentimentIntensityAnalyzer()
54
-
55
- def get_sentiment(stock="HBL"):
56
- if not NEWS_KEY:
57
- return 0
58
- try:
59
- url = f"https://newsapi.org/v2/everything?q={stock}+Pakistan&apiKey={NEWS_KEY}"
60
- r = requests.get(url).json()
61
- if "articles" not in r:
62
- return 0
63
- articles = r["articles"][:5]
64
- scores = [analyzer.polarity_scores(a["title"])['compound'] for a in articles]
65
- return np.mean(scores) if scores else 0
66
- except:
67
- return 0
68
-
69
- # ------------------------------
70
- # Predict Next Day
71
  # ------------------------------
72
  def predict_next(df):
73
- data = scaler.fit_transform(df[["Close"]])
74
- last60 = data[-60:].reshape(1, 60, 1)
75
- pred = model.predict(last60, verbose=0)[0][0]
76
- pred_real = scaler.inverse_transform([[pred]])[0][0]
77
- return pred_real
 
 
 
 
78
 
79
  # ------------------------------
80
  # Streamlit UI
81
  # ------------------------------
82
- st.title("πŸ“ˆ PSX Stock Predictor – FYP Version")
83
- st.write("Live PSX price trend + sentiment + ML prediction")
84
-
85
  symbol = st.selectbox("Choose PSX Stock:", ["HBL", "UBL", "ENGRO", "PSO", "OGDC"])
86
 
87
  if st.button("Fetch & Predict"):
88
- with st.spinner("Fetching data..."):
89
- df = get_psx_data(symbol)
90
- sentiment = get_sentiment(symbol)
91
  prediction = predict_next(df)
92
 
93
- # Adjust prediction with sentiment
94
- sentiment_adj = prediction + (prediction * sentiment * 0.02)
95
-
96
  # Plot historical + predicted
97
  fig, ax = plt.subplots()
98
  ax.plot(df.index, df["Close"], label="Historical Price")
99
- ax.axhline(sentiment_adj, linestyle="--", label="Predicted Price")
100
  ax.set_title(f"{symbol} Stock Price Prediction")
101
  ax.legend()
102
  st.pyplot(fig)
103
 
104
- # Display results
105
  st.subheader("Prediction Result")
106
- st.write(f"**Predicted Price:** Rs {sentiment_adj:.2f}")
107
- st.write(f"**Sentiment Impact:** {sentiment:.3f}")
 
1
  import streamlit as st
2
  import numpy as np
3
  import pandas as pd
 
 
 
 
 
 
4
  import matplotlib.pyplot as plt
5
  import os
6
 
7
+ # Optional imports for ML, but safely handled
 
 
8
  try:
9
+ import tensorflow as tf
10
+ from tensorflow.keras.models import load_model
11
+ from sklearn.preprocessing import MinMaxScaler
12
+ import joblib
13
+ except ImportError:
14
+ tf = None
15
+ load_model = None
16
+ MinMaxScaler = None
17
+ joblib = None
18
+
19
+ st.set_page_config(page_title="PSX Stock Predictor", layout="wide")
20
 
21
+ st.title("πŸ“ˆ PSX Stock Predictor – HF Safe Version")
22
+ st.write("This app works even if model or API is missing.")
23
 
24
  # ------------------------------
25
+ # Load Model & Scaler
26
  # ------------------------------
27
+ MODEL_LOADED = False
28
+ if tf and os.path.exists("model.h5"):
29
+ try:
30
+ model = load_model("model.h5", custom_objects={"mse": tf.keras.metrics.MeanSquaredError()})
31
+ scaler = joblib.load("scaler.pkl") if os.path.exists("scaler.pkl") else MinMaxScaler()
32
+ MODEL_LOADED = True
33
+ st.success("Model loaded successfully!")
34
+ except Exception as e:
35
+ st.warning(f"Model found but failed to load: {e}")
36
+ else:
37
+ st.warning("Model not found. Using dummy predictions.")
 
 
 
 
38
 
39
+ # ------------------------------
40
+ # Dummy / Fallback Data
41
+ # ------------------------------
42
+ def get_dummy_data():
43
  dates = pd.date_range(end=pd.Timestamp.today(), periods=200)
44
  prices = np.linspace(100, 150, 200) + np.random.normal(0, 2, 200)
45
  df = pd.DataFrame({"Close": prices}, index=dates)
46
  return df
47
 
48
  # ------------------------------
49
+ # Prediction function
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
50
  # ------------------------------
51
  def predict_next(df):
52
+ if MODEL_LOADED:
53
+ data = scaler.fit_transform(df[["Close"]])
54
+ last60 = data[-60:].reshape(1, 60, 1)
55
+ pred = model.predict(last60, verbose=0)[0][0]
56
+ pred_real = scaler.inverse_transform([[pred]])[0][0]
57
+ return pred_real
58
+ else:
59
+ # Dummy prediction: next value + small random change
60
+ return df["Close"].iloc[-1] * (1 + np.random.uniform(-0.01, 0.01))
61
 
62
  # ------------------------------
63
  # Streamlit UI
64
  # ------------------------------
 
 
 
65
  symbol = st.selectbox("Choose PSX Stock:", ["HBL", "UBL", "ENGRO", "PSO", "OGDC"])
66
 
67
  if st.button("Fetch & Predict"):
68
+ with st.spinner("Fetching data and predicting..."):
69
+ df = get_dummy_data()
 
70
  prediction = predict_next(df)
71
 
 
 
 
72
  # Plot historical + predicted
73
  fig, ax = plt.subplots()
74
  ax.plot(df.index, df["Close"], label="Historical Price")
75
+ ax.axhline(prediction, linestyle="--", color="red", label="Predicted Price")
76
  ax.set_title(f"{symbol} Stock Price Prediction")
77
  ax.legend()
78
  st.pyplot(fig)
79
 
80
+ # Display result
81
  st.subheader("Prediction Result")
82
+ st.write(f"**Predicted Price:** Rs {prediction:.2f}")