Modules for Import¶

In [1]:
import pandas as pd

# !pip install pandasai
# !pip install langchain_community
# !pip install langchain_openai

import os
# from pandasai import Agent
from pandasai import SmartDataframe
from pandasai.llm import OpenAI
# from pandasai.helpers.openai_info import get_openai_callback
import logging
import openai

# from langchain_openai.chat_models import ChatOpenAI

# Create a custom logging configuration
logging.basicConfig(level=logging.CRITICAL)

Read NBA Stats Data CSV from disk¶

In [2]:
# Load your data into a pandas DataFrame
df = pd.read_csv('nba-stats-data.csv', low_memory=False)
# set season (year) as main DF index to allow for time series analysis of player data
df = df.reset_index().set_index('season')
df.index = pd.to_datetime(df.index, format='%Y').to_period('Y')
df = df.sort_index(axis=0)
df.head()
Out[2]:
index assists blocks comment defReb fga fgm fgp fta ftm ... DOB age feet meters position active jersey_number years_pro rookie_year pounds
season
2015 34 0.0 1.0 NaN 1.0 3.0 2.0 66.7 2.0 1.0 ... 1991-07-01 33.0 6"10 2.08 F-C True 33.0 8.0 2013.0 240.0
2015 35 0.0 1.0 NaN 1.0 3.0 2.0 66.7 2.0 1.0 ... 1991-07-01 33.0 6"10 2.08 F-C True 33.0 8.0 2013.0 240.0
2015 36 2.0 0.0 NaN 7.0 7.0 3.0 42.9 2.0 2.0 ... 1988-07-16 36.0 6"7 2.01 F True 1.0 9.0 2012.0 237.0
2015 37 0.0 0.0 NaN 3.0 4.0 0.0 0.0 0.0 0.0 ... NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN
2015 38 1.0 1.0 NaN 5.0 5.0 3.0 60.0 6.0 2.0 ... 1992-03-22 33.0 7"3 2.21 C False 40.0 1.0 2015.0 265.0

5 rows × 45 columns

Analysis: Speak to the Data!¶

In [29]:
df.columns
Out[29]:
Index(['index', 'assists', 'blocks', 'comment', 'defReb', 'fga', 'fgm', 'fgp',
       'fta', 'ftm', 'ftp', 'min', 'offReb', 'pFouls', 'plusMinus', 'points',
       'pos', 'steals', 'totReb', 'tpa', 'tpm', 'tpp', 'turnovers',
       'player_id', 'player_name', 'team_code', 'team_name', 'team_id',
       'game_id', 'affiliation', 'college', 'firstname', 'id', 'lastname',
       'country', 'DOB', 'age', 'feet', 'meters', 'position', 'active',
       'jersey_number', 'years_pro', 'rookie_year', 'pounds'],
      dtype='object')

Set Open AI API Key¶

In [48]:
os.environ["OPENAI_API_KEY"] = "*******"

llm = OpenAI(model="gpt-3.5-turbo", openai_api_key=openai.api_key)

Set Large Language Model & other DF confirgurations¶

In [49]:
df_nba = SmartDataframe(df, config={"llm": llm, "conversational": True, "verbose": True})

Chat with (query data)¶

In [6]:
df_nba.chat("Which are the top 20 player_names by assists?")
Out[6]:
'The top 20 players by assists are: Rajon Rondo, Russell Westbrook, Russell Westbrook, Russell Westbrook, Russell Westbrook, Tyrese Haliburton, Russell Westbrook, Russell Westbrook, Trae Young, Rajon Rondo, Russell Westbrook, Russell Westbrook, Russell Westbrook, Russell Westbrook, Russell Westbrook, James Harden, Russell Westbrook, Russell Westbrook, Elfrid Payton, Russell Westbrook'
In [7]:
df_nba.chat("Which are the top 10 player_names by assists?")
Out[7]:
player_name
season
2017 Rajon Rondo
2018 Russell Westbrook
2020 Russell Westbrook
2018 Russell Westbrook
2020 Russell Westbrook
2023 Tyrese Haliburton
2016 Russell Westbrook
2016 Russell Westbrook
2024 Trae Young
2017 Rajon Rondo
In [8]:
df_nba.chat("Which are the top 5 player_names group by points?")
Out[8]:
points
player_name
Kevin Durant 35170.0
Russell Westbrook 33398.0
Kyrie Irving 29574.0
Andre Drummond 27846.0
Nikola Vucevic 27172.0
In [9]:
df_nba.chat("Which are the top 5 player_names group by blocks?")
Out[9]:
player_name blocks
761 JaVale McGee 2648.0
78 Andre Drummond 2478.0
1351 Mason Plumlee 1866.0
1162 Kevin Durant 1640.0
1662 Rudy Gobert 1586.0
In [10]:
df_nba.chat("Which are the top 5 player_names group by assists?")
Out[10]:
player_name assists
1664 Russell Westbrook 13532.0
1209 Kyle Lowry 9276.0
593 Eric Bledsoe 7404.0
841 James Harden 6952.0
1215 Kyrie Irving 6658.0
In [11]:
df_nba.chat("Which are the top 5 distinct player_names by pounds?")
Out[11]:
player_name pounds
season
2015 Boban Marjanovic 290.0
2015 Jusuf Nurkic 290.0
2015 Nikola Jokic 284.0
2019 Zion Williamson 284.0
2015 Joel Embiid 280.0
In [12]:
df_nba.chat("Which are the top 5 distinct player_names by meters?")
Out[12]:
player_name meters
season
2015 Edy Tavares 2.21
2015 Kristaps Porzingis 2.21
2015 Boban Marjanovic 2.21
2019 Bol Bol 2.18
2015 Rudy Gobert 2.16
In [13]:
df_nba.chat("Which are the top 5 player_names group by plusMinus?")
Out[13]:
plusMinus
player_name
Kevin Durant 8420.0
Stephen Curry 5453.0
Kyrie Irving 5234.0
Danny Green 5004.0
Kyle Lowry 4852.0
In [14]:
df_nba.chat("Which are the top 5 player_names by all stats excluding pFouls and turnovers?")

<string>:2: FutureWarning: The default value of numeric_only in DataFrameGroupBy.sum is deprecated. In a future version, numeric_only will default to False. Either specify numeric_only or select only columns which should be valid for the function.
<string>:2: FutureWarning: The default value of numeric_only in DataFrameGroupBy.sum is deprecated. In a future version, numeric_only will default to False. Either specify numeric_only or select only columns which should be valid for the function.
Out[14]:
index assists blocks defReb fga fgm fgp fta ftm ftp ... player_id team_id game_id id age meters jersey_number years_pro rookie_year pounds
player_name
Russell Westbrook 373101044 13532.0 534.0 10396.0 27712.0 12152.0 68476.8 9478.0 7020.0 104525.6 ... 874752.0 35742 11163800 874752.0 57888.0 3055.20 0.0 20904.0 3228864.0 321600.0
Kyle Lowry 481096796 9276.0 530.0 5582.0 16670.0 7122.0 61007.8 5470.0 4556.0 94909.6 ... 500964.0 49560 10451772 500964.0 59748.0 2803.56 10724.0 22980.0 3073192.0 300272.0
Eric Bledsoe 376362837 7404.0 654.0 4980.0 17730.0 7965.0 65280.6 5640.0 4407.0 96492.3 ... 93102.0 35598 8487396 93102.0 55230.0 2919.30 7890.0 17358.0 3171780.0 337692.0
James Harden 121821995 6952.0 515.0 4408.0 14255.0 6214.0 34177.2 6897.0 5959.0 67193.1 ... 175392.0 11937 5622842 175392.0 28420.0 1591.52 812.0 9744.0 1631308.0 178640.0
Kyrie Irving 66193858 6658.0 582.0 4054.0 22846.0 11004.0 57969.4 4868.0 4374.0 95042.6 ... 328860.0 6586 8781624 328860.0 41580.0 2368.80 13860.0 12600.0 2533860.0 245700.0

5 rows × 29 columns

In [15]:
df_nba.chat("Which are the best players by single game stats?")

<string>:2: SettingWithCopyWarning: 
A value is trying to be set on a copy of a slice from a DataFrame.
Try using .loc[row_indexer,col_indexer] = value instead

See the caveats in the documentation: https://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy
Out[15]:
player_name assists blocks defReb fga fgm fta ftm min offReb plusMinus points steals totReb turnovers total_score
season
2023 Giannis Antetokounmpo 3.0 1.0 11.0 28.0 20.0 32.0 24.0 37.0 3.0 29.0 64.0 4.0 14.0 4.0 233.0
2023 Joel Embiid 5.0 1.0 9.0 41.0 24.0 23.0 21.0 37.0 9.0 11.0 70.0 1.0 18.0 1.0 233.0
2018 James Harden 4.0 0.0 9.0 38.0 17.0 25.0 22.0 NaN 6.0 19.0 61.0 5.0 15.0 5.0 221.0
2022 Donovan Mitchell 11.0 1.0 5.0 34.0 22.0 25.0 20.0 50.0 3.0 19.0 71.0 0.0 8.0 4.0 219.0
2022 Joel Embiid 8.0 7.0 10.0 28.0 19.0 24.0 20.0 37.0 1.0 25.0 59.0 1.0 11.0 5.0 213.0
In [50]:
df_nba.chat("Plot points scored by LeBron James showing for each season, using different colors for each bar")
Out[50]:
'/Users/conorheffron/workspace/data-science-in-python/assignment-1/submission/v2/exports/charts/temp_chart.png'
No description has been provided for this image
No description has been provided for this image
In [17]:
df_nba.chat("Plot heatmap of the best 7 player_names")

<string>:2: FutureWarning: The default value of numeric_only in DataFrameGroupBy.sum is deprecated. In a future version, numeric_only will default to False. Either specify numeric_only or select only columns which should be valid for the function.
Out[17]:
'/Users/conorheffron/workspace/data-science-in-python/assignment-1/submission/v2/exports/charts/temp_chart.png'
No description has been provided for this image
In [ ]: