Introduction¶
A dashboard of forecasts of the Presidential Election for 2020.
Visual can be found here: https://public.tableau.com/profile/michael.dunphy8764#!/vizhome/2020Elections-USPresident/MainDashboard?publish=yes
Data Source: https://en.wikipedia.org/wiki/2020_United_States_presidential_election
Below is the code used to collect and modify the data to produce the Tableau visual.
In [1]:
!pip install lxml
!pip install html5lib
import requests
import pandas as pd
from bs4 import BeautifulSoup
In [2]:
# Access Wiki page with forcasting data from Cook, IE, Sabato, Politico, RCP, Niskanen, CNN,
# The Economist, CBS News, 270toWin, ABC News, NPR, NBC News, 538, CNalysis
r = requests.get('https://en.wikipedia.org/wiki/2020_United_States_presidential_election')
root = BeautifulSoup(r.content)
find = root.find('table', class_="wikitable sortable")
li = pd.read_html(find.prettify())
ratings = pd.DataFrame(li[0])
# Ratings will be the main data set cleaned and used for the visual
ratings.head()
Out[2]:
In [3]:
# Data table of dates the columns were last updated on Wiki
updated = pd.DataFrame(ratings.columns.values[4:])
updated.to_csv("updated.csv", index= False)
updated.head()
Out[3]:
In [4]:
# Data Cleaning
ratings.columns = ["State", "Electoral Votes", "PVI", "2016 result", "Average Ratings", "Cook", "IE", "Sabato",
"Politico", "RCP", "Niskanen", "CNN", "The Economist", "CBS", "270toWin", "ABC", "NPR", "NBC",
"538", "CNalysis"]
ratings.drop(ratings.index[-1:], inplace= True)
ratings.head()
Out[4]:
In [5]:
# Used to count the number of groupings for each available state
def get_count(row, prediction):
count = 0
column_names = ratings.columns.values[5:19]
for col in column_names:
if prediction in row[col]:
count += 1
return count
In [6]:
# Used to get consensus ruling for the state of the race for that state
def get_consensus(row):
m = 0
column_names = ratings.columns.values[20:30]
for col in column_names:
if row[col] >= m:
m = ratings.iloc[index][col]
prediction = col
return prediction, m
In [7]:
# Adds numbers of groups to each state
safe_r = []
likely_r = []
lean_r = []
tilt_r = []
toss = []
tilt_d = []
lean_d = []
likely_d = []
safe_d = []
for index, row in ratings.iterrows():
safe_r.append(get_count(row, "Safe R"))
likely_r.append(get_count(row, "Likely R"))
lean_r_c = 0
lean_r_c += (get_count(row, "Leans R"))
lean_r_c += (get_count(row, "Lean R"))
lean_r.append(lean_r_c)
tilt_r.append(get_count(row, "Tilt R"))
toss.append(get_count(row, "Tossup"))
safe_d.append(get_count(row, "Safe D"))
likely_d.append(get_count(row, "Likely D"))
lean_d.append(get_count(row, "Lean D"))
tilt_d.append(get_count(row, "Tilt D"))
ratings["Safe R"] = safe_r
ratings["Likely R"] = likely_r
ratings["Lean R"] = lean_r
ratings["Tilt R"] = tilt_r
ratings["Tossup"] = toss
ratings["Safe D"] = safe_d
ratings["Likely D"] = likely_d
ratings["Lean D"] = lean_d
ratings["Tilt D"] = tilt_d
ratings.head()
Out[7]:
In [8]:
# Adds consensus ruling, max number of same groupings, and confidence of consensus ruling
prediction = []
m = []
for index, row in ratings.iterrows():
p, em = get_consensus(row)
prediction.append(p)
m.append(em)
ratings["Consensus"] = prediction
ratings["Max"] = m
ratings["Confidence"] = ratings["Max"] / 15
ratings.head()
Out[8]:
In [9]:
# Combines data to be used in visual
combined = ratings
combined.to_csv('combined.csv', index = False)
combined.head()
Out[9]: