Build a Free Cloud AI Speed‑Dating Model with Alibaba PAI‑DSW

Artificial intelligence (AI) has evolved from its inception in the 1950s to become a dominant technology, with machine learning as its core sub‑field. Deep learning, powered by frameworks like TensorFlow and GPUs, has driven breakthroughs in vision, speech, and recommendation systems.

Setting up a local deep‑learning environment can be cumbersome due to OS, GPU, driver, and library dependencies. Alibaba Cloud’s PAI‑DSW (Data Science Workshop) offers a cloud‑based IDE that integrates JupyterLab, WebIDE, and terminal access, providing free GPU resources for personal developers.

Getting Started with PAI‑DSW

Visit https://dsw-dev.data.aliyun.com/#/ and log in with an Alibaba Cloud account.

Create a new DSW lab; the interface shows a file explorer, workspace, and resource panel.

Open a Python 3 notebook; a new .ipynb file is created.

Write and run code by selecting a cell and pressing Shift+Enter.

Example: print a welcome message.

print("Welcome to DSW 👏👏👏 You can use this lab for any data experiment you like 😝")

Next, compute the 10th Fibonacci number.

# Compute the Nth Fibonacci number
def Fibonacci(n):
    if n < 0:
        print("Invalid input, please provide a positive integer")
    elif n == 1:
        return 0
    elif n == 2:
        return 1
    else:
        return Fibonacci(n-1) + Fibonacci(n-2)

print('The 10th Fibonacci number is:', Fibonacci(10))

Case Study: Predicting Speed‑Dating Success

The dataset contains ~8,000 rapid‑dating rounds with features such as gender, age, attractiveness, sincerity, intelligence, and match outcome.

import pandas as pd
df = pd.read_csv('Speed Dating Data.csv', encoding='gbk')
print(df.shape)

Exploratory analysis shows only 16.47% of rounds result in a match. Gender‑wise success rates differ slightly, with females having a marginally higher match probability.

# Plot gender match rates
import matplotlib.pyplot as plt
size_of_groups = df[df.gender == 0].match.value_counts().values
single_pct = round(size_of_groups[0]/sum(size_of_groups)*100, 2)
matched_pct = round(size_of_groups[1]/sum(size_of_groups)*100, 2)
labels = [f'Single: {single_pct}%', f'Matched: {matched_pct}%']
plt.pie(size_of_groups, labels=labels)
plt.show()

Age distribution peaks between 22‑28 years, suggesting most participants are young adults.

# Age histogram
import numpy as np
age = df[np.isfinite(df['age'])]['age']
plt.hist(age, bins=35)
plt.xlabel('Age')
plt.ylabel('Frequency')
plt.show()

Correlation analysis highlights that attributes such as attractiveness, sincerity, intelligence, fun, ambition, and shared interests are strongly related to the match outcome.

# Correlation heatmap
import seaborn as sns
selected = df[['attr_o','sinc_o','intel_o','fun_o','amb_o','shar_o','match']]
corr = selected.corr()
sns.heatmap(corr, xticklabels=corr.columns, yticklabels=corr.columns)
plt.show()

Modeling steps:

Select the six predictor columns and the target match.

Drop rows with missing values.

Apply SVMSMOTE to balance the classes.

Split into training (80%) and test (20%) sets.

Train a logistic regression model.

# Data preparation
clean_df = df[['attr_o','sinc_o','intel_o','fun_o','amb_o','shar_o','match']].dropna()
X = clean_df[['attr_o','sinc_o','intel_o','fun_o','amb_o','shar_o']]
y = clean_df['match']

# Balance classes
from imblearn.over_sampling import SVMSMOTE
X_res, y_res = SVMSMOTE().fit_resample(X, y)

# Train‑test split
from sklearn.model_selection import train_test_split
X_train, X_test, y_train, y_test = train_test_split(X_res, y_res, test_size=0.2, random_state=0, stratify=y_res)

# Logistic regression
from sklearn.linear_model import LogisticRegression
from sklearn import metrics
model = LogisticRegression(C=1, random_state=0)
model.fit(X_train, y_train)
train_acc = metrics.accuracy_score(y_train, model.predict(X_train))
test_acc = metrics.accuracy_score(y_test, model.predict(X_test))
print('Training Accuracy:', train_acc)
print('Validation Accuracy:', test_acc)

The balanced logistic regression achieves around 83% validation accuracy, demonstrating a viable approach to predict whether a speed‑dating round will result in a match.