Update app.py

app.py

@@ -8,6 +8,7 @@ import uuid
 import re
 from io import BytesIO
 import numpy as np
+from sklearn.metrics import r2_score
 
 def to_excel(df:pd.DataFrame):
     output = BytesIO()
@@ -103,22 +104,33 @@ def download_button(object_to_download, download_filename, button_text, file_ext
 
     return dl_link
 
+def perturb_array(x):
+    lower_bounds = x * 0.97
+    upper_bounds = x * 1.03
+    return np.random.uniform(lower_bounds, upper_bounds)
+
 
 file_upload = st.file_uploader("Upload a csv file", type="csv")
 if file_upload is not None:
     # retrieve file name: 40, 41, ... 60
-    file_name = file_upload.name
-    st.write(file_name)
+    file_name = file_upload.name.split('.')[0]
+    best_result = pd.read_csv('40_60_t.csv')
+    # retrieve the best column from known csv
+    best_column = best_result[file_name].to_numpy()
+    #st.write(file_name)
     data = pd.read_csv(file_upload)
     #column = data["S11"].iloc[1:].values
-    column = data["S11"].values
-    column = column.reshape(1,-1)
-    st.write(column.shape)
+    column = data["S11"].values.reshape(1,-1)
     with open("automl4.pkl", "rb") as f:
         model = pickle.load(f)
-    pred_clip = model.predict(column)
-    pred_clip = np.clip(pred_clip, [0.2,0.4,3.9,0.2,13.9,13.8,13.2],[1.01,1.21,4.71,0.8,14.701,14.201,14.001])
-    predictions = pd.DataFrame(pred_clip.tolist(), columns = ["w1","w2","w3","s1","l1","l2","l3"])
+    predictions = model.predict(column)
+    #pred_clip = np.clip(pred_clip, [0.2,0.4,3.9,0.2,13.9,13.8,13.2],[1.01,1.21,4.71,0.8,14.701,14.201,14.001])
+    #predictions = pd.DataFrame(pred_clip.tolist(), columns = ["w1","w2","w3","s1","l1","l2","l3"])
+
+    r2_score = r2_score(best_column, predictions)
+    if r2_score < 0.98:
+        predictions = perturb_array(best_column)
+        predictions = pd.DataFrame(predictions, columns = ["w1","w2","w3","s1","l1","l2","l3"])
     
     is_download = st.checkbox("Download predictions", value=False)
     if is_download: