Titanic

Python source code	Jupyter Notebook	Task
titanic.py	titanic.ipynb	Machine Learning 📈

Titanic is a famous playground competition hosted by Kaggle (here), so I'll simply copy-paste its brief description here:

This is the legendary Titanic ML competition – the best, first challenge for you to dive into ML competitions and familiarize yourself with how the Kaggle platform works.

The competition is simple: use machine learning to create a model that predicts which passengers survived the Titanic shipwreck.

Here are the frist few rows of the train.csv of Titanic:

PassengerId,Survived,Pclass,Name,Sex,Age,SibSp,Parch,Ticket,Fare,Cabin,Embarked

1,0,3,"Braund, Mr. Owen Harris",male,22,1,0,A/5 21171,7.25,,S

2,1,1,"Cumings, Mrs. John Bradley (Florence Briggs Thayer)",female,38,1,0,PC 17599,71.2833,C85,C

3,1,3,"Heikkinen, Miss. Laina",female,26,0,0,STON/O2. 3101282,7.925,,S

And the first few rows of the test.csv:

PassengerId,Pclass,Name,Sex,Age,SibSp,Parch,Ticket,Fare,Cabin,Embarked

892,3,"Kelly, Mr. James",male,34.5,0,0,330911,7.8292,,Q

893,3,"Wilkes, Mrs. James (Ellen Needs)",female,47,1,0,363272,7,,S

894,2,"Myles, Mr. Thomas Francis",male,62,0,0,240276,9.6875,,Q

What we need to do is to predict the Survived column in test.csv.

# preparations

import torch

import cflearn

import numpy as np

# for reproduction

np.random.seed(142857)

torch.manual_seed(142857)

Configurations

Since the target column is not the last column (which is the default setting of carefree-learn), we need to manually configure it:

kwargs = dict(carefree=True, cf_data_config={"label_name": "Survived"})

And you're all set! Notice that only the label_name needs to be provided, and carefree-learn will find out the corresponding target column for you😉

Build Your Model

For instance, we'll use the famous Wide & Deep model. Unlike other libraries, carefree-learn supports file-in:

m = cflearn.api.fit_ml("train.csv", core_name="wnd", **kwargs)

Evaluate Your Model

After building the model, we can directly evaluate our model with a file (file-out):

# instantiate an `MLInferenceData` instance

idata = cflearn.MLInferenceData("train.csv")

# `contains_labels` is set to True because we're evaluating on training set

cflearn.ml.evaluate(idata, metrics=["acc", "auc"], pipelines=m, contains_labels=True)

Then you will see something like this:

> [ info ] Results

================================================================================================================================

| metrics | acc | auc |

--------------------------------------------------------------------------------------------------------------------------------

| | mean | std | score | mean | std | score |

--------------------------------------------------------------------------------------------------------------------------------

| wnd | 0.857463 | 0.000000 | 0.857463 | 0.892550 | 0.000000 | 0.892550 |

================================================================================================================================

Our model achieved an accuracy of 0.857463, not bad!

info

Note that this performance may vary and is not exactly the training performance, because carefree-learn will automatically split out the cross validation dataset for you.

Making Predictions

Again, we can directly make predictions with a file (file-out):

# instantiate an `MLInferenceData` instance

idata = cflearn.MLInferenceData("test.csv")

# `contains_labels` is set to False because `test.csv` does not contain labels

predictions = m.predict(idata, make_loader_kwargs={"contains_labels": False})

Submit Your Results

If you reached here, we have actually already completed this Titanic task! All we need to do is to convert the predictions into a submission file:

def write_submissions(name: str, predictions_: np.ndarray) -> None:

with open("test.csv", "r") as f:

f.readline()

id_list = [line.strip().split(",")[0] for line in f]

with open(name, "w") as f:

f.write("PassengerId,Survived\n")

for test_id, prediction in zip(id_list, predictions_):

f.write(f"{test_id},{prediction}\n")

predictions = predictions[cflearn.PREDICTIONS_KEY]

write_submissions("submissions.csv", predictions.argmax(1))

After running these codes, a submissions.csv will be generated and you can submit it to Kaggle directly. In my personal experience, it could achieve 0.77272.

Conclusions

Since Titanic is just a small toy dataset, using Neural Network to solve it might actually 'over-killed' (or, overfit) it, and that's why we decided to conclude here instead of introducing more fancy techniques (e.g. ensemble, AutoML, etc.). We hope that this small example can help you quickly walk through some basic concepts in carefre-learn, as well as help you leverage carefree-learn in your own tasks!