I write models for cases on Kaggle , study strangers and get inspired. All articles describing how to implement them in a web project, for me, a junior frontend student , give an overhead of complex information, but I just want to "borrow" any cool model and quickly implement it into my service. Hands itched to come up with a universal algorithm, so a solution was found quickly.
Let's get started. Step 1
I want to take the format of most models from Kaggle , so that in the future it would be easy to borrow someone else's code of any complexity without understanding it. For Boat carts write in the Python 3.9 , using liby pyTelegramBotAPI , to address the compatibility of extensions .py and .ipynb yuzaem ipynb .
And so, we install the dependencies:
pip install pyTelegramBotAPI pip install ipynb
Go to Kaggle and choose the model you like. I'll start with the classic Titanic case - Machine Learning from Disaster , borrow this solution (Titanic Random Forest: 82.78%), drag it into the bot project.
Install the appeared dependencies:
pip install <>
We saw the bot. Step 2
Create a new file and import our libs into it:
import telebot
from ipynb.fs.defs.ml import is_user_alive
ipynb.fs.defs.ml
ml , , is_user_alive
, . , .
, ( @BotFather):
bot = telebot.TeleBot('token')
/start
, . , .
@bot.messagehandler(commands=['start'])
def welcome(message):
bot.sendmessage(message.chat.id, '! , ( ), '
' (male/female), , (1-, 0-), '
' (1-, 0-), , '
', - , !')
, . :
@bot.messagehandler(contenttypes=['text'])
def answer(message):
bot.sendmessage(message.chat.id, '…')
passengerdata = message.text.split()
passengerdata.insert(0, 0)
passengerdata.insert(9, ',')
passengerdata[2] = '"', passengerdata[2], '"'
, , 9 , .
. 3
cntrl+f
#%%
. . :
<>
def is_user_alive(user_data):
< >
- , :
with open(os.path.join('input', 'test.csv'), "a") as fp:
wr = csv.writer(fp, dialect='excel')
wr.writerow(user_data)
Predictions
( ). , (return
):
return predictions[len(predictions)-1:]['Survived']
. 4
answer
, 0 1… . , :
answer = is_user_alive(passenger_data)
if int(answer) == 1:
bot.sendmessage(message.chat.id, '! , .')
elif int(answer) == 0:
bot.send_message(message.chat.id, ', … .')
We create a function with a proposal to check someone else, call it from answer
with an message
argument:
def doagain(message):
bot.sendmessage(message.chat.id, ' - ?')
We start polling:
while True:
try:
bot.polling(none_stop=True)
except ():
time.sleep(5)
Result
Everything! Pretty simple, right?
If not, you can watch the video version:
Code: https://github.com/freakssha/ml-bot-titanic
This is a speedrun for itchy hands, it is not optimized and it can be improved a lot, but I don’t know how yet. If you understand how to do this without losing simplicity and versatility - write, please!