Netflix analisis with help of OpenAI

Created by Darío López Díaz. Work in progress..

In this notebook, we use chatgpt 3.5 to analize data from a netflix titles dataframe. The idea is to give prompts to chatgpt directly from python and directly apply the response of chatgpt in order to compute different interesting things.

# OpenAI ChatGPT LLM. import openai import os import pandas as pd import matplotlib.pyplot as plt from dotenv import load_dotenv, find_dotenv _ = load_dotenv(find_dotenv())

We define the get_ completition function, with the variables defined by, model: gpt-3.5-turbo and temperature=0 that defines at the minimum, the level of randomness (level of "creativity") of the model. In this way the results provided from the model are more exact.

def get_completion(prompt, model="gpt-3.5-turbo"): messages = [{"role": "user", "content": prompt}] response = openai.ChatCompletion.create( model=model, messages=messages, temperature=0, # this is the degree of randomness of the model's output ) return response.choices[0].message["content"] # We charge the DataFrame. df = pd.read_csv('archive/netflix_titles.csv') df.head()

Number of shows by country / countries.

df['country'].value_counts()

We ask to ChatGPT to do the same, it will create the program and also implement it in Python. From chatGPT we ask to compute the number of shows by country / countries.

text_1 = f"""{df}""" prompt = f""" You will be provided with a dataframe delimited by triple quotes. / Give me a code in pandas to compute what is the number of shows by country. / Sort the result starting by the biggest number. / Don't use comments. Do not import packages in the code. \"\"\"{text_1}\"\"\" """ response = get_completion(prompt) print(response) eval(response)

So, in this way ChatGPT can do almost everything. It will write and apply the program in Python. For example, to show what is the most repeated actor in the data and show some information about the actor.

text_1 = f"""{df}""" prompt1 = f""" You will be provided with a dataframe delimited by triple quotes. / Give me a code in pandas to compute what is the most repeated actor in the cast column, and the number of times that the actor appears. / Don't use comments. Do not import packages in the code. \"\"\"{text_1}\"\"\" """ response1 = eval(str(get_completion(prompt1))) prompt2 = f""" Tell me information you might know (not from the dataset) of the actor from the first two strings of \"\"\"{response1}\"\"\" """ response2 = get_completion(prompt2) print(response1) print(response2)

Now the most repeated director in the data.

text_1 = f"""{df}""" prompt1 = f""" You will be provided with a dataframe delimited by triple quotes. / Give me a code in pandas to compute what is the most repeated director in the director column, and the number of times that the director appears. / Don't use comments. Do not import packages in the code. \"\"\"{text_1}\"\"\" """ response1 = eval(get_completion(prompt1)) prompt2 = f""" Tell me information you might know (not from the dataset) of the director from the first two strings of \"\"\"{response1}\"\"\" """ response2 = get_completion(prompt2) print(response1) print(response2)

We ask to GPT to do some plots, for example: a pie plot with the percentage of type of shows.

text_1 = f"""{df}""" prompt = f""" You will be provided with a dataframe delimited by triple quotes. / Give me a code in pandas to pie plot the percentage of the diferent types of shows from the 'type' column, with respect to the total of shows. / Include the plot labels Movie and Tv Show, where Movie correspond to the biggest percentage. / Don't use comments. Do not import packages in the code. \"\"\"{text_1}\"\"\" """ response = get_completion(prompt) print(response) eval(response)

Next, we plot the number of shows by year, from 1990.

text_1 = f"""{df}""" prompt = f""" You will be provided with a dataframe delimited by triple quotes. / Give me a code in pandas to plot the number of shows, with respect to the 'release_year' column, starting from the year '1990'. / Don't use comments. Do not import packages in the code. \"\"\"{text_1}\"\"\" """ response = get_completion(prompt) eval(response)

The maximum of shows is at 2019. Obviously, the Pandemic affected the productions of shows since 2019.