experiment

functions/calc_matches.py

@@ -53,6 +53,32 @@ def calc_matches(filtered_df, project_df, similarity_matrix, top_x):
     # Select submatrix based on indices from both dataframes
     match_matrix = similarity_matrix[filtered_df_indices, :][:, project_df_indices]
 
+    st.write(match_matrix.shape)
+
+    flattened_indices = np.argsort(match_matrix, axis=None)[-15:]
+
+    # Step 2: Convert flattened indices to 2D indices
+    row_indices, col_indices = np.unravel_index(flattened_indices, match_matrix.shape)
+
+    # Step 3: Extract the top 15 values and their corresponding indices
+    top_values = match_matrix[row_indices, col_indices]
+    top_indices = list(zip(row_indices, col_indices, top_values))
+
+    # Step 4: Sort the indices and values based on the values in descending order
+    top_15_indices_sorted = sorted(top_indices, key=lambda x: x[2], reverse=True)
+
+    # Display the results
+    for idx, (row, col, value) in enumerate(top_15_indices_sorted):
+        st.write(f"Rank {idx + 1}: Value = {value}, Row Index = {row}, Column Index = {col}")
+
+    p1_df = filtered_df.iloc[row_indices].copy()
+    p1_df['similarity'] = top_values
+    p2_df = project_df.iloc[col_indices].copy()
+    p2_df['similarity'] = top_values
+
+    return p1_df, p2_df
+
+    """
     # Get the linear indices of the top 'top_x' values
     # (flattened index to handle the sparse matrix more effectively)
     linear_indices = np.argsort(match_matrix.data)[-top_x:]
@@ -74,4 +100,5 @@ def calc_matches(filtered_df, project_df, similarity_matrix, top_x):
     print("finished calc matches")
 
     return p1_df, p2_df
+    """