Create app.py

app.py

@@ -0,0 +1,149 @@
+import pandas as pd
+import joblib
+from huggingface_hub import HfApi
+import pickle
+import yfinance as yf
+from datetime import datetime, timedelta
+from forex_python.converter import get_rate
+import pandas as pd
+import numpy as np
+import cpi
+from sklearn.preprocessing import MinMaxScaler
+
+#cpi.update()
+
+import gradio as gr
+
+from huggingface_hub import notebook_login
+notebook_login()
+import hopsworks
+
+# from huggingface_hub import hf_hub_download
+# m = hf_hub_download(repo_id="marvmk/model-test", filename="model.pkl")
+
+project = hopsworks.login(api_key_value="4CY1rwa8iz8Yu6gG.TwayrYmsX4GQfhSp3LNKYTLvyFMfqAvnzNUQp5ae9K5HhfYxb5mcnLAutm1K18zV")
+mr = project.get_model_registry()
+model = mr.get_model("stock_modal", version=1)
+model_dir = model.download()
+model = joblib.load(model_dir + "/stock_model.pkl")
+
+#model = pickle.load(open(m, 'rb'))
+
+# downloading the last 10 days to make the prediction
+from datetime import date
+
+today = date.today()
+days_ago = today - timedelta(days=20)
+
+# we get the last 20 days and keep just the last 10 working days, which have prices
+nasdaq = yf.Ticker("^IXIC")
+hist = nasdaq.history(start=days_ago, end=today)
+hist = hist.drop(columns=['Dividends', 'Stock Splits'])
+
+# keeping the last 10 data points
+hist = hist[-10:]
+
+
+inflation = []
+for t in hist.index:
+  inflation.append(get_rate("USD", "EUR", t))
+
+cpi_items_df = cpi.series.get(seasonally_adjusted=False).to_dataframe()
+cpi_items_df = cpi_items_df[cpi_items_df['period_type']=='monthly']
+cpi_items_df['date'] = pd.to_datetime(cpi_items_df['date'])
+cpi_items_df = cpi_items_df.set_index('date')
+cpi_df = cpi_items_df['value'].loc['2022':'2023']
+
+cpi_col = []
+for x in hist.index:
+  # ts = datetime(x.year, x.month, 1)
+
+  # just adding the latest inflation rate
+  cpi_col.append(cpi_df[-1])
+
+hist['Inflation'] = inflation
+hist['CPI'] = cpi_col
+
+hist
+
+hist['Quarter_end'] = np.where(hist.index.month%3==0,1,0)
+
+# def build_sequences(nump, target_labels=['Close'], window=10, stride=1, telescope=1):
+#     # Sanity check to avoid runtime errors
+#     df = pd.DataFrame(nump, columns = ['Open','High','Low','Close','Volume','Inflation', 'CPI', 'Quarter_end'])
+#     assert window % stride == 0
+#     dataset = []
+#     labels = []
+#     temp_df = df.copy().values
+#     temp_label = df[target_labels].copy().values
+#     padding_len = len(df)%window
+
+#     if(padding_len != 0):
+#         # Compute padding length
+#         padding_len = window - len(df)%window
+#         padding = np.zeros((padding_len,temp_df.shape[1]), dtype='float64')
+#         temp_df = np.concatenate((padding,df))
+#         padding = np.zeros((padding_len,temp_label.shape[1]), dtype='float64')
+#         temp_label = np.concatenate((padding,temp_label))
+#         assert len(temp_df) % window == 0
+
+#     for idx in np.arange(0,len(temp_df)-window-telescope,stride):
+#         dataset.append(temp_df[idx:idx+window])
+#         labels.append(temp_label[idx+window:idx+window+telescope])
+
+#     dataset = np.array(dataset)
+#     labels = np.array(labels)
+#     return dataset, labels
+
+hist
+
+s = hf_hub_download(repo_id="marvmk/scalable_project", filename="scaler.save", repo_type='dataset')
+scaler = joblib.load(s)
+
+inp = scaler.transform(hist.to_numpy())
+#inp = scaler.inverse_transform(inp)
+
+df = inp
+temp_df = pd.DataFrame(inp, columns = ['Open','High','Low','Close','Volume','Inflation', 'CPI', 'Quarter_end'])
+ds = []
+ds.append(temp_df[0:10])
+ds = np.array(ds)
+
+ds
+
+predictions = model.predict(ds)
+predictions
+p = predictions[0][0]
+print(p)
+a = np.array([0,0,0,p,0,0,0,0])
+
+a = scaler.inverse_transform(a.reshape(1,-1))
+a
+
+final_prediction = a[-1][3]
+
+final_prediction
+
+import matplotlib.pyplot as plt
+import streamlit as st
+
+prediction = []
+#prediction.append(final_prediction)
+close = hist['Close'].to_list()
+print(close)
+for c in close:
+  prediction.append(c)
+
+
+prediction.append(final_prediction)
+print(prediction)
+plt.figure(figsize = (20,10))
+plt.plot(prediction, label="Prediction")
+plt.plot(hist['Close'].to_list()[-10:], label="Previous")
+plt.ylabel('Price US$', fontsize = 15 )
+plt.xlabel('Working Days', fontsize = 15 )
+plt.title("NASDAQ Stock Prediction", fontsize = 20)
+plt.legend()
+plt.grid()
+
+st.pyplot(plt)