如何解决统计模型中 MN Logit 回归中的优势比
我有这个由 statsmodel 完成的 Multi Numinal 回归模型:
writer = pd.ExcelWriter(path=os.path.join(export_path,f'regression.xlsx'),engine='xlsxwriter')
vars_matrix_df = pd.read_csv(data_path,skipinitialspace=True)
corr_cols = ['sales_vs_service','agent_experience','minutes_passed_since_shift_started','stage_in_conv','current_cust_wait_time','prev_cust_line_words','total_cust_words_in_conv','agent_total_turns','sentiment_score','max_sentiment','min_sentiment','last_sentiment','agent_response_time','customer_response_rate','is_last_cust_answered','conversation_opening','queue_length','total_lines_from_rep','agent_number_of_conversations','concurrency','rep_shift_start_time','first_cust_line_num_of_words','queue_wait_time','day_of_week','time_of_day']
reg_equation = st.formula.mnlogit(f'visitor_was_answered ~C(day_of_week)+C(time_of_day)+{"+".join(corr_cols)} ',vars_matrix_df).fit()
注册结果:
visitor_was_answered=1 coef std err z P>|z| \
0 C(time_of_day)[T.10] 0.0071 1910000.000 3.700000e-09 1.000
1 C(time_of_day)[T.11] 0.0067 698000.000 9.600000e-09 1.000
2 C(time_of_day)[T.12] 0.0016 1790000.000 9.200000e-10 1.000
3 C(time_of_day)[T.13] 0.0031 561000.000 5.570000e-09 1.000
4 C(time_of_day)[T.14] 0.0037 1310000.000 2.840000e-09 1.000
5 C(time_of_day)[T.15] 0.0011 548000.000 2.020000e-09 1.000
6 C(time_of_day)[T.17] 0.0044 814000.000 5.440000e-09 1.000
7 C(time_of_day)[T.18] 0.0009 1100000.000 8.270000e-10 1.000
8 C(time_of_day)[T.19] 0.0047 835000.000 5.640000e-09 1.000
9 C(time_of_day)[T.20] 0.0009 1140000.000 8.100000e-10 1.000
10 time_of_day[T.10] 0.0071 1930000.000 3.670000e-09 1.000
11 time_of_day[T.11] 0.0067 686000.000 9.770000e-09 1.000
12 time_of_day[T.12] 0.0016 1800000.000 9.150000e-10 1.000
13 time_of_day[T.13] 0.0031 556000.000 5.620000e-09 1.000
14 time_of_day[T.14] 0.0037 1240000.000 3.010000e-09 1.000
15 time_of_day[T.15] 0.0011 638000.000 1.740000e-09 1.000
16 time_of_day[T.17] 0.0044 1010000.000 4.400000e-09 1.000
17 time_of_day[T.18] 0.0009 1130000.000 8.020000e-10 1.000
18 time_of_day[T.19] 0.0047 860000.000 5.480000e-09 1.000
19 time_of_day[T.20] 0.0009 1120000.000 8.270000e-10 1.000
20 sales_vs_service -0.0448 0.006 -8.102000e+00 0.000
21 agent_experience -0.0414 0.008 -4.955000e+00 0.000
22 current_cust_wait_time -39.1333 0.414 -9.457400e+01 0.000
23 prev_cust_line_words 20.0439 0.236 8.494600e+01 0.000
24 agent_total_turns 0.1110 0.038 2.949000e+00 0.003
25 sentiment_score -4.3454 0.157 -2.759000e+01 0.000
26 agent_response_time -118.0821 2.205 -5.354600e+01 0.000
27 customer_response_rate -7.0865 0.630 -1.125500e+01 0.000
28 is_last_cust_answered -0.2537 0.005 -4.860800e+01 0.000
29 conversation_opening -0.4533 0.006 -7.206300e+01 0.000
30 queue_length -1.5427 0.018 -8.642700e+01 0.000
31 agent_number_of_conversations 0.0013 0.018 7.300000e-02 0.941
32 first_cust_line_num_of_words -3.7545 0.123 -3.056900e+01 0.000
33 queue_wait_time -0.3308 0.166 -1.997000e+00 0.046
对于这个回归,我想添加每个变量的优势比值。我认为系数已经是优势比,但我没有找到任何证据。知道如何做到这一点吗?这里的系数是什么?
谢谢!
版权声明:本文内容由互联网用户自发贡献,该文观点与技术仅代表作者本人。本站仅提供信息存储空间服务,不拥有所有权,不承担相关法律责任。如发现本站有涉嫌侵权/违法违规的内容, 请发送邮件至 dio@foxmail.com 举报,一经查实,本站将立刻删除。