
#Import Libraries
import pandas as pd
import numpy as np
from pandas.tseries.offsets import BDay
#Allows us to filter out non business days


# Install a pip package in the current Jupyter kernel
import sys
!{sys.executable} -m pip install plotly
import plotly.express as px

Requirement already satisfied: plotly in /Users/samchilds/opt/anaconda3/lib/python3.8/site-packages (4.14.3)
Requirement already satisfied: retrying>=1.3.3 in /Users/samchilds/opt/anaconda3/lib/python3.8/site-packages (from plotly) (1.3.3)
Requirement already satisfied: six in /Users/samchilds/opt/anaconda3/lib/python3.8/site-packages (from plotly) (1.15.0)


#S&P 500 Tweets:
spyDF = pd.read_csv("twitterData/tweetsSPY.csv")
del spyDF["Unnamed: 0"]
#Import file, remove unnecessary index column
spyDF["Datetime"] = pd.to_datetime(spyDF["Datetime"], format = "%m/%d/%Y %H:%M")
#Convert existing datetime column from text to datetime object type


#Tesla Tweets:
tslaDF = pd.read_csv("twitterData/tweetsTSLA.csv")
del tslaDF["Unnamed: 0"]
tslaDF["Datetime"] = pd.to_datetime(tslaDF["Datetime"], format = "%m/%d/%Y %H:%M")


#Gamestop Tweets:
gmeDF = pd.read_csv("twitterData/tweetsGME.csv")
del gmeDF["Unnamed: 0"]
gmeDF["Datetime"] = pd.to_datetime(gmeDF["Datetime"], format = "%m/%d/%Y %H:%M")


#Exxon Mobil Tweets:
xomDF = pd.read_csv("twitterData/tweetsXOM.csv")
del xomDF["Unnamed: 0"]
xomDF["Datetime"] = pd.to_datetime(xomDF["Datetime"], format = "%m/%d/%Y %H:%M")


#SPY Price Data
stonkSPY = pd.read_excel("priceData/SPY.xlsx")
stonkSPY["Datetime"] = pd.to_datetime(stonkSPY["Date"], format = "%Y-%m-%d")
#Before importing with this statement, the price data from yahoo finance needs to be formatted to match pandas.
del stonkSPY["Date"]
stonkSPY.set_index('Datetime')


#TSLA Price Data
stonkTSLA = pd.read_excel("priceData/TSLA.xlsx")
stonkTSLA["Datetime"] = pd.to_datetime(stonkTSLA["Date"], format = "%Y-%m-%d")
del stonkTSLA["Date"]
stonkTSLA.set_index('Datetime')


#GME Price Data
stonkGME = pd.read_excel("priceData/GME.xlsx")
stonkGME["Datetime"] = pd.to_datetime(stonkGME["Date"], format = "%Y-%m-%d")
del stonkGME["Date"]
stonkGME.set_index('Datetime')


#XOM Price Data
stonkXOM = pd.read_excel("priceData/XOM.xlsx")
stonkXOM["Datetime"] = pd.to_datetime(stonkXOM["Date"], format = "%Y-%m-%d")
del stonkXOM["Date"]
stonkXOM.set_index('Datetime')


#SPY Sentiment
spyDF = spyDF.drop('TweetID',1).resample('d', on="Datetime").mean()
spyDF = spyDF.dropna()
spyDF2 = spyDF.reset_index()
validSPY = spyDF2["Datetime"].tolist()
spySentiment = pd.Series(spyDF2["Predictions"].values, index = spyDF2["Datetime"]).to_dict()
#Creates a dictionary with the date time object as key and the day's sentiment average as the value.


#TSLA Sentiment
tslaDF = tslaDF.drop('TweetID',1).resample('d', on="Datetime").mean()
tslaDF = tslaDF.dropna()
tslaDF2 = tslaDF.reset_index()
validTSLA = tslaDF2["Datetime"].tolist()
tslaSentiment = pd.Series(tslaDF2["Predictions"].values, index = tslaDF2["Datetime"]).to_dict()


#GME Sentiment
gmeDF = gmeDF.drop('TweetID',1).resample('d', on="Datetime").mean()
gmeDF = gmeDF.dropna()
gmeDF2 = gmeDF.reset_index()
validGME = gmeDF2["Datetime"].tolist()
gmeSentiment = pd.Series(gmeDF2["Predictions"].values, index = gmeDF2["Datetime"]).to_dict()


#XOM Sentiment
xomDF = xomDF.drop('TweetID',1).resample('d', on="Datetime").mean()
xomDF = xomDF.dropna()
xomDF2 = xomDF.reset_index()
validXOM = xomDF2["Datetime"].tolist()
xomSentiment = pd.Series(xomDF2["Predictions"].values, index = xomDF2["Datetime"]).to_dict()


#SPY
rotDF = stonkSPY.drop(["Open","High","Low","Close","Adj Close","Volume"],axis = 1)
rotDF = rotDF.set_index("Datetime")
meltDF = rotDF.reset_index()
meltDF2 = meltDF.melt(id_vars = "Datetime")
meltDF3 = meltDF2.set_index("Datetime").sort_values(by=['Datetime','variable'])
meltDF4 = meltDF3.rename(columns={"variable":"holdingPeriod",'value':'return'})
meltDF4 = meltDF4.reset_index()
meltDF5 = meltDF4[meltDF4["Datetime"].isin(validSPY)]

newTable = []
for row in meltDF5.itertuples(index = False):
    comp = list(row)
    comp.append(spyDF.loc[row[0]]["Predictions"])
    newTable.append(tuple(comp))

df6 = pd.DataFrame(newTable, columns = ["date", "holdP","return","sentiment"])
finalSPY = df6.dropna()


#TSLA
rotDF = stonkTSLA.drop(["Open","High","Low","Close","Adj Close","Volume"],axis = 1)
rotDF = rotDF.set_index("Datetime")
meltDF = rotDF.reset_index()
meltDF2 = meltDF.melt(id_vars = "Datetime")
meltDF3 = meltDF2.set_index("Datetime").sort_values(by=['Datetime','variable'])
meltDF4 = meltDF3.rename(columns={"variable":"holdingPeriod",'value':'return'})
meltDF4 = meltDF4.reset_index()
meltDF5 = meltDF4[meltDF4["Datetime"].isin(validTSLA)]

newTable = []
for row in meltDF5.itertuples(index = False):
    comp = list(row)
    comp.append(tslaDF.loc[row[0]]["Predictions"])
    newTable.append(tuple(comp))

df6 = pd.DataFrame(newTable, columns = ["date", "holdP","return","sentiment"])
finalTSLA = df6.dropna()


#GME
rotDF = stonkGME.drop(["Open","High","Low","Close","Adj Close","Volume"],axis = 1)
rotDF = rotDF.set_index("Datetime")
meltDF = rotDF.reset_index()
meltDF2 = meltDF.melt(id_vars = "Datetime")
meltDF3 = meltDF2.set_index("Datetime").sort_values(by=['Datetime','variable'])
meltDF4 = meltDF3.rename(columns={"variable":"holdingPeriod",'value':'return'})
meltDF4 = meltDF4.reset_index()
meltDF5 = meltDF4[meltDF4["Datetime"].isin(validGME)]

newTable = []
for row in meltDF5.itertuples(index = False):
    comp = list(row)
    comp.append(gmeDF.loc[row[0]]["Predictions"])
    newTable.append(tuple(comp))

df6 = pd.DataFrame(newTable, columns = ["date", "holdP","return","sentiment"])
finalGME = df6.dropna()


#XOM
rotDF = stonkXOM.drop(["Open","High","Low","Close","Adj Close","Volume"],axis = 1)
rotDF = rotDF.set_index("Datetime")
meltDF = rotDF.reset_index()
meltDF2 = meltDF.melt(id_vars = "Datetime")
meltDF3 = meltDF2.set_index("Datetime").sort_values(by=['Datetime','variable'])
meltDF4 = meltDF3.rename(columns={"variable":"holdingPeriod",'value':'return'})
meltDF4 = meltDF4.reset_index()
meltDF5 = meltDF4[meltDF4["Datetime"].isin(validXOM)]

newTable = []
for row in meltDF5.itertuples(index = False):
    comp = list(row)
    comp.append(xomDF.loc[row[0]]["Predictions"])
    newTable.append(tuple(comp))

df6 = pd.DataFrame(newTable, columns = ["date", "holdP","return","sentiment"])
finalXOM = df6.dropna()


#Create Visuals
lbls = {"sentiment": "Sentiment Score",
          "holdP" : "Time Horizon",
          "return": "Investment Return"
}

#*** Just get the dataframes updated and our work here is done.

ttl = "SPY Returns by Holding Period and Simple Average Sentiment\n" 
spyfig = px.scatter_3d(finalSPY, x="sentiment", y="holdP", z="return", 
                    hover_name="date", color="holdP",range_color = [0,12], 
                    range_x=[0.43,0.29], title = ttl,labels = lbls)
spyfig.show()

ttl = "TSLA Returns by Holding Period and Simple Average Sentiment\n" 
tslafig = px.scatter_3d(finalTSLA, x="sentiment", y="holdP", z="return", 
                    hover_name="date", color="holdP",range_color = [0,12], 
                    range_x=[0.8,0.67], title = ttl,labels = lbls)
tslafig.show()

ttl = "GME Returns by Holding Period and Simple Average Sentiment\n" 
gmefig = px.scatter_3d(finalGME, x="sentiment", y="holdP", z="return", 
                    hover_name="date", color="holdP",range_color = [0,12], 
                    range_x=[0.7,0.4], title = ttl,labels = lbls)
gmefig.show()

ttl = "XOM Returns by Holding Period and Simple Average Sentiment\n" 
xomfig = px.scatter_3d(finalXOM, x="sentiment", y="holdP", z="return", 
                    hover_name="date", color="holdP",range_color = [0,12], 
                    range_x=[0.85,0.6], title = ttl,labels = lbls)
xomfig.show()

	Open	High	Low	Close	Adj Close	Volume	1	2	3	4	5	10
Datetime
2020-04-16	40.480000	40.480000	38.930000	39.150002	36.160999	30934000	0.103959	0.051852	0.046232	0.076118	0.109834	0.186973
2020-04-17	39.880001	43.320000	39.869999	43.220001	39.920265	40829100	-0.047200	-0.052291	-0.025220	0.005322	0.011800	-0.001851
2020-04-20	40.750000	42.740002	40.540001	41.180000	38.036011	39260800	-0.005342	0.023070	0.055124	0.061923	0.067023	0.089850
2020-04-21	40.400002	42.130001	39.630001	40.959999	37.832809	44381600	0.028565	0.060791	0.067627	0.072754	0.097900	0.094483
2020-04-22	42.599998	42.930000	41.860001	42.130001	38.913486	27084700	0.031332	0.037978	0.042962	0.067410	0.126513	0.044149
...	...	...	...	...	...	...	...	...	...	...	...	...
2021-04-09	56.000000	56.439999	55.529999	55.869999	55.869999	20010100	-0.006622	-0.002864	0.025953	0.019868	NaN	NaN
2021-04-12	56.759998	56.790001	55.310001	55.500000	55.500000	20238800	0.003784	0.032793	0.026667	NaN	NaN	NaN
2021-04-13	55.299999	55.980000	55.090000	55.709999	55.709999	19370600	0.028900	0.022797	NaN	NaN	NaN	NaN
2021-04-14	56.410000	58.009998	56.250000	57.320000	57.320000	28051100	-0.005932	NaN	NaN	NaN	NaN	NaN
2021-04-15	57.419998	57.419998	56.779999	56.980000	56.980000	23246300	NaN	NaN	NaN	NaN	NaN	NaN

By Sam Childs and Taoyu Yang¶

Import Libraries for Data Processing and Graphing¶

Import Tweets¶

Import Price Data¶

Calculate Average Daily Sentiment¶

Combine Daily Sentiment Averages and Price Data¶

Generate our Visuals¶

Does Social Media Sentiment Predict Stock Price¶

By Sam Childs and Taoyu Yang¶

Import Libraries for Data Processing and Graphing¶

Import Tweets¶

Import Price Data¶

Calculate Average Daily Sentiment¶

Combine Daily Sentiment Averages and Price Data¶

Generate our Visuals¶