analysis

analysis_sum_result.py

@@ -0,0 +1,185 @@
+from orm import engine
+import pandas as pd
+import networkx as nx
+import json
+import matplotlib.pyplot as plt
+
+# prep data
+Firm = pd.read_csv("Firm_amended.csv")
+Firm['Code'] = Firm['Code'].astype('string')
+Firm.fillna(0, inplace=True)
+BomNodes = pd.read_csv('BomNodes.csv', index_col=0)
+
+with open('SQL_analysis_risk.sql', 'r') as f:
+    str_sql = f.read()
+result = pd.read_sql(sql=str_sql,
+                     con=engine)
+result.to_csv('analysis\\count.csv',
+              index=False,
+              encoding='utf-8-sig')
+print(result)
+
+# G bom
+plt.rcParams['font.sans-serif'] = 'SimHei'
+
+exp_id = 1
+G_bom_str = pd.read_sql(
+    sql=f'select g_bom from iiabmdb.without_exp_experiment '
+        f'where id = {exp_id};',
+    con=engine)['g_bom'].tolist()[0]
+G_bom = nx.adjacency_graph(json.loads(G_bom_str))
+pos = nx.nx_agraph.graphviz_layout(G_bom, prog="twopi", args="")
+node_labels = nx.get_node_attributes(G_bom, 'Name')
+plt.figure(figsize=(12, 12), dpi=300)
+nx.draw_networkx_nodes(G_bom, pos)
+nx.draw_networkx_edges(G_bom, pos)
+nx.draw_networkx_labels(G_bom, pos, labels=node_labels, font_size=6)
+# plt.show()
+plt.savefig(f"analysis\\g_bom_exp_id_{exp_id}.png")
+plt.close()
+
+# G firm
+plt.rcParams['font.sans-serif'] = 'SimHei'
+
+sample_id = 1
+G_firm_str = pd.read_sql(
+    sql=f'select g_firm from iiabmdb.without_exp_sample where id = {exp_id};',
+    con=engine)['g_firm'].tolist()[0]
+G_firm = nx.adjacency_graph(json.loads(G_firm_str))
+pos = nx.nx_agraph.graphviz_layout(G_firm, prog="twopi", args="")
+node_label = nx.get_node_attributes(G_firm, 'Name')
+# desensitize
+node_label = {
+    key: key
+    for key in node_label.keys()
+}
+node_size = list(nx.get_node_attributes(G_firm, 'Revenue_Log').values())
+node_size = list(map(lambda x: x**2, node_size))
+edge_label = nx.get_edge_attributes(G_firm, "Product")
+edge_label = {(n1, n2): label for (n1, n2, _), label in edge_label.items()}
+plt.figure(figsize=(12, 12), dpi=300)
+nx.draw(G_firm, pos, node_size=node_size, labels=node_label, font_size=6)
+nx.draw_networkx_edge_labels(G_firm, pos, edge_label, font_size=4)
+# plt.show()
+plt.savefig(f"analysis\\g_firm_sample_id_{exp_id}_de.png")
+plt.close()
+
+# count firm product
+count_firm_prod = result.value_counts(subset=['id_firm', 'id_product'])
+count_firm_prod.name = 'count'
+count_firm_prod = count_firm_prod.to_frame().reset_index()
+count_firm_prod = pd.merge(count_firm_prod,
+                           Firm[['Code', 'Name']],
+                           how='left',
+                           left_on='id_firm',
+                           right_on='Code')
+count_firm_prod.drop('Code', axis=1, inplace=True)
+count_firm_prod.rename(columns={'Name': 'name_firm'}, inplace=True)
+count_firm_prod = pd.merge(count_firm_prod,
+                           BomNodes[['Code', 'Name']],
+                           how='left',
+                           left_on='id_product',
+                           right_on='Code')
+count_firm_prod.drop('Code', axis=1, inplace=True)
+count_firm_prod.rename(columns={'Name': 'name_product'}, inplace=True)
+count_firm_prod = count_firm_prod[[
+    'id_firm', 'name_firm', 'id_product', 'name_product', 'count'
+]]
+count_firm_prod.to_csv('analysis\\count_firm_prod.csv',
+                       index=False,
+                       encoding='utf-8-sig')
+print(count_firm_prod)
+
+# count firm
+count_firm = count_firm_prod.groupby('id_firm')['count'].sum()
+count_firm = count_firm.to_frame().reset_index()
+count_firm = pd.merge(count_firm,
+                      Firm[['Code', 'Name']],
+                      how='left',
+                      left_on='id_firm',
+                      right_on='Code')
+count_firm.drop('Code', axis=1, inplace=True)
+count_firm.sort_values('count', inplace=True, ascending=False)
+count_firm = count_firm[['id_firm', 'Name', 'count']]
+count_firm.to_csv('analysis\\count_firm.csv',
+                  index=False,
+                  encoding='utf-8-sig')
+print(count_firm)
+
+# count product
+count_prod = count_firm_prod.groupby('id_product')['count'].sum()
+count_prod = count_prod.to_frame().reset_index()
+count_prod = pd.merge(count_prod,
+                      BomNodes[['Code', 'Name']],
+                      how='left',
+                      left_on='id_product',
+                      right_on='Code')
+count_prod.drop('Code', axis=1, inplace=True)
+count_prod.sort_values('count', inplace=True, ascending=False)
+count_prod = count_prod[['id_product', 'Name', 'count']]
+count_prod.to_csv('analysis\\count_prod.csv',
+                  index=False,
+                  encoding='utf-8-sig')
+print(count_prod)
+
+# DCP disruption causing probability
+result_disrupt_ts_above_0 = result[result['ts'] > 0]
+print(result_disrupt_ts_above_0)
+result_dcp = pd.DataFrame(columns=[
+    's_id', 'up_id_firm', 'up_id_product', 'down_id_firm', 'down_id_product'
+])
+for sid, group in result.groupby('s_id'):
+    ts_start = max(group['ts'])
+    while ts_start >= 1:
+        ts_end = ts_start - 1
+        while ts_end >= 0:
+            up = group.loc[group['ts'] == ts_end, ['id_firm', 'id_product']]
+            down = group.loc[group['ts'] == ts_start,
+                             ['id_firm', 'id_product']]
+            for _, up_row in up.iterrows():
+                for _, down_row in down.iterrows():
+                    row = [sid]
+                    row += up_row.tolist()
+                    row += down_row.tolist()
+                    result_dcp.loc[len(result_dcp.index)] = row
+            ts_end -= 1
+        ts_start -= 1
+count_dcp = result_dcp.value_counts(
+    subset=['up_id_firm', 'up_id_product', 'down_id_firm', 'down_id_product'])
+count_dcp.name = 'count'
+count_dcp = count_dcp.to_frame().reset_index()
+count_dcp = pd.merge(count_dcp,
+                     Firm[['Code', 'Name']],
+                     how='left',
+                     left_on='up_id_firm',
+                     right_on='Code')
+count_dcp.drop('Code', axis=1, inplace=True)
+count_dcp.rename(columns={'Name': 'up_name_firm'}, inplace=True)
+count_dcp = pd.merge(count_dcp,
+                     BomNodes[['Code', 'Name']],
+                     how='left',
+                     left_on='up_id_product',
+                     right_on='Code')
+count_dcp.drop('Code', axis=1, inplace=True)
+count_dcp.rename(columns={'Name': 'up_name_product'}, inplace=True)
+count_dcp = pd.merge(count_dcp,
+                     Firm[['Code', 'Name']],
+                     how='left',
+                     left_on='down_id_firm',
+                     right_on='Code')
+count_dcp.drop('Code', axis=1, inplace=True)
+count_dcp.rename(columns={'Name': 'down_name_firm'}, inplace=True)
+count_dcp = pd.merge(count_dcp,
+                     BomNodes[['Code', 'Name']],
+                     how='left',
+                     left_on='down_id_product',
+                     right_on='Code')
+count_dcp.drop('Code', axis=1, inplace=True)
+count_dcp.rename(columns={'Name': 'down_name_product'}, inplace=True)
+count_dcp = count_dcp[[
+    'up_id_firm', 'up_name_firm', 'up_id_product', 'up_name_product',
+    'down_id_firm', 'down_name_firm', 'down_id_product', 'down_name_product',
+    'count'
+]]
+count_dcp.to_csv('analysis\\count_dcp.csv', index=False, encoding='utf-8-sig')
+print(count_dcp)