загрузка данных¶

In [5]:

import pandas as pd
df = pd.read_csv("C://Users//annal//aim//static//csv//Forbes_Billionaires.csv")
df

Out[5]:

	Rank	Name	Networth	Age	Country	Source	Industry
0	1	Elon Musk	219.0	50	United States	Tesla, SpaceX	Automotive
1	2	Jeff Bezos	171.0	58	United States	Amazon	Technology
2	3	Bernard Arnault & family	158.0	73	France	LVMH	Fashion & Retail
3	4	Bill Gates	129.0	66	United States	Microsoft	Technology
4	5	Warren Buffett	118.0	91	United States	Berkshire Hathaway	Finance & Investments
...	...	...	...	...	...	...	...
2595	2578	Jorge Gallardo Ballart	1.0	80	Spain	pharmaceuticals	Healthcare
2596	2578	Nari Genomal	1.0	82	Philippines	apparel	Fashion & Retail
2597	2578	Ramesh Genomal	1.0	71	Philippines	apparel	Fashion & Retail
2598	2578	Sunder Genomal	1.0	68	Philippines	garments	Fashion & Retail
2599	2578	Horst-Otto Gerberding	1.0	69	Germany	flavors and fragrances	Food & Beverage

2600 rows × 7 columns

входные и выходные переменные:¶

входные: age и country¶

выходные: networth¶

создание лингвистических переменных¶

In [6]:

import skfuzzy as fuzz
from skfuzzy import control as ctrl
age_cat = ctrl.Antecedent(df["Age"].sort_values(), "age_cat")
networth = ctrl.Antecedent(df["Networth"].sort_values(), "networth")
rank = ctrl.Consequent(df["Rank "].sort_values(), "rank")

формирование нечетких переменных для лингвистических переменных и их визуализация¶¶

In [7]:

age_cat.automf(3, variable_type= "quant")
age_cat.view()
networth.automf(3, variable_type="quant")
networth.view()
rank.automf(5, variable_type="quant")
rank.view()

c:\Users\annal\aim\.venv\Lib\site-packages\skfuzzy\control\fuzzyvariable.py:125: UserWarning: FigureCanvasAgg is non-interactive, and thus cannot be shown
  fig.show()

No description has been provided for this image

формирование и визуализация нечётких правил¶

In [13]:

import skfuzzy as fuzz
from skfuzzy import control as ctrl 
rule1 = ctrl.Rule(age_cat['low'] & networth['low'], rank['lower'])
rule2 = ctrl.Rule(age_cat['low'] & networth['average'], rank['low'])
rule3 = ctrl.Rule(age_cat['low'] & networth['high'], rank['average'])
rule4 = ctrl.Rule(age_cat['average'] & networth['low'], rank['lower'])
rule5 = ctrl.Rule(age_cat['average'] & networth['average'], rank['average'])
rule6 = ctrl.Rule(age_cat['average'] & networth['high'], rank['high'])
rule7 = ctrl.Rule(age_cat['high'] & networth['low'], rank['average'])
rule8 = ctrl.Rule(age_cat['high'] & networth['average'], rank['high'])
rule9 = ctrl.Rule(age_cat['high'] & networth['high'], rank['higher'])

rule1.view()

Out[13]:

(<Figure size 640x480 with 1 Axes>, <Axes: >)

создание нечеткой системы и добавление нечетких правил в базу знаний нечеткой системы¶

In [14]:

rank_ctrl = ctrl.ControlSystem(
    [
        rule1,
        rule2,
        rule3,
        rule4,
        rule5,
        rule6,
        rule7,
        rule8,
        rule9,
    ]
)

ranks = ctrl.ControlSystemSimulation(rank_ctrl)

пример расчёта rank на основе age_cat и networth¶

In [15]:

ranks.input['age_cat'] = 50
ranks.input['networth'] = 219
ranks.compute()
ranks.print_state()
print(ranks.output['rank'])

=============
 Antecedents 
=============
Antecedent: age_cat                 = 50
  - low                             : 0.2345679012345679
  - average                         : 0.7654320987654321
  - high                            : 0.0
Antecedent: networth                = 219
  - low                             : 0.0
  - average                         : 0.0
  - high                            : 1.0

=======
 Rules 
=======
RULE #0:
  IF age_cat[low] AND networth[low] THEN rank[lower]
	AND aggregation function : fmin
	OR aggregation function  : fmax

  Aggregation (IF-clause):
  - age_cat[low]                                           : 0.2345679012345679
  - networth[low]                                          : 0.0
                            age_cat[low] AND networth[low] = 0.0
  Activation (THEN-clause):
                                               rank[lower] : 0.0

RULE #1:
  IF age_cat[low] AND networth[average] THEN rank[low]
	AND aggregation function : fmin
	OR aggregation function  : fmax

  Aggregation (IF-clause):
  - age_cat[low]                                           : 0.2345679012345679
  - networth[average]                                      : 0.0
                        age_cat[low] AND networth[average] = 0.0
  Activation (THEN-clause):
                                                 rank[low] : 0.0

RULE #2:
  IF age_cat[low] AND networth[high] THEN rank[average]
	AND aggregation function : fmin
	OR aggregation function  : fmax

  Aggregation (IF-clause):
  - age_cat[low]                                           : 0.2345679012345679
  - networth[high]                                         : 1.0
                           age_cat[low] AND networth[high] = 0.2345679012345679
  Activation (THEN-clause):
                                             rank[average] : 0.2345679012345679

RULE #3:
  IF age_cat[average] AND networth[low] THEN rank[lower]
	AND aggregation function : fmin
	OR aggregation function  : fmax

  Aggregation (IF-clause):
  - age_cat[average]                                       : 0.7654320987654321
  - networth[low]                                          : 0.0
                        age_cat[average] AND networth[low] = 0.0
  Activation (THEN-clause):
                                               rank[lower] : 0.0

RULE #4:
  IF age_cat[average] AND networth[average] THEN rank[average]
	AND aggregation function : fmin
	OR aggregation function  : fmax

  Aggregation (IF-clause):
  - age_cat[average]                                       : 0.7654320987654321
  - networth[average]                                      : 0.0
                    age_cat[average] AND networth[average] = 0.0
  Activation (THEN-clause):
                                             rank[average] : 0.0

RULE #5:
  IF age_cat[average] AND networth[high] THEN rank[high]
	AND aggregation function : fmin
	OR aggregation function  : fmax

  Aggregation (IF-clause):
  - age_cat[average]                                       : 0.7654320987654321
  - networth[high]                                         : 1.0
                       age_cat[average] AND networth[high] = 0.7654320987654321
  Activation (THEN-clause):
                                                rank[high] : 0.7654320987654321

RULE #6:
  IF age_cat[high] AND networth[low] THEN rank[average]
	AND aggregation function : fmin
	OR aggregation function  : fmax

  Aggregation (IF-clause):
  - age_cat[high]                                          : 0.0
  - networth[low]                                          : 0.0
                           age_cat[high] AND networth[low] = 0.0
  Activation (THEN-clause):
                                             rank[average] : 0.0

RULE #7:
  IF age_cat[high] AND networth[average] THEN rank[high]
	AND aggregation function : fmin
	OR aggregation function  : fmax

  Aggregation (IF-clause):
  - age_cat[high]                                          : 0.0
  - networth[average]                                      : 0.0
                       age_cat[high] AND networth[average] = 0.0
  Activation (THEN-clause):
                                                rank[high] : 0.0

RULE #8:
  IF age_cat[high] AND networth[high] THEN rank[higher]
	AND aggregation function : fmin
	OR aggregation function  : fmax

  Aggregation (IF-clause):
  - age_cat[high]                                          : 0.0
  - networth[high]                                         : 1.0
                          age_cat[high] AND networth[high] = 0.0
  Activation (THEN-clause):
                                              rank[higher] : 0.0


==============================
 Intermediaries and Conquests 
==============================
Consequent: rank                     = 1756.4538971994773
  lower:
    Accumulate using accumulation_max : 0.0
  low:
    Accumulate using accumulation_max : 0.0
  average:
    Accumulate using accumulation_max : 0.2345679012345679
  high:
    Accumulate using accumulation_max : 0.7654320987654321
  higher:
    Accumulate using accumulation_max : 0.0

1756.4538971994773

тестирование нечёткой системы¶

In [16]:

def fuzzy_pred(row):
    ranks.input['age_cat'] = row['Age']
    ranks.input['networth'] = row['Networth']
    ranks.compute()
    return ranks.output['rank']

res = df[['Age', 'Networth', 'Rank ']].head(100)

res['Pred'] = res.apply(fuzzy_pred, axis=1)

res.head(15)

Out[16]:

	Age	Networth	Rank	Pred
0	50	219.0	1	1756.453897
1	58	171.0	2	1604.640029
2	73	158.0	3	1622.578064
3	66	129.0	4	1441.738174
4	91	118.0	5	1768.807248
5	49	111.0	6	1106.377953
6	48	107.0	7	1083.278263
7	77	106.0	8	1541.073986
8	66	91.4	9	1283.707578
9	64	90.7	10	1245.266049
10	59	90.0	11	1143.677808
11	80	82.0	12	1412.560349
12	82	81.2	13	1429.370589
13	68	74.8	14	1227.428422
14	37	67.3	15	889.877260

оценка результатов (метрики для задачи регрессии)¶

In [17]:

import math
from sklearn import metrics


rmetrics = {}
rmetrics["RMSE"] = math.sqrt(metrics.mean_squared_error(res['Rank '], res['Pred']))
rmetrics["RMAE"] = math.sqrt(metrics.mean_absolute_error(res['Rank '], res['Pred']))
rmetrics["R2"] = metrics.r2_score(res['Rank '], res['Pred'])

rmetrics

Out[17]:

{'RMSE': 1028.1501824418408,
 'RMAE': 31.426056535415803,
 'R2': -1272.901233304339}

p.s. менять датасет нет смысла. все подобные датасеты имеют одинаковые колонки. какой не возьми, резульат не хороший

207 KiB Raw Blame History Unescape Escape