Mini projects

Churn prediction

Binary Classification Classic ML Tabular
Notebook
The task is to predict churn of telecom company customers.
Implementation includes:

• Exploratory Data Analysis
  • Missing Values Imputation
  • Categorical Columns Encoding
  • Data Normalization
• Modeling:
  • Baseline: SimpleImputer + StandardScaler + OneHotEncoder + Logistic regression + GridSearchCV
  • Final: CatBoostClassifier + Hyperparameter Search (Optuna)

Salary prediction

Regression DL NLP
Notebook
The task is to predict salary based on the different text and categorical features.
Implementation includes:

• Exploratory Data Analysis
  • Categorical Columns Encoding
  • Target transformation
• Modeling:
  • Baseline: Custom PyTorch dataset + Custom Transforms + Fusion model (Title Encoder + Description Encoder + Categorical Encoder)
  • Improved model: In progress
• Explaining model predictions: In progress

Simpsons classification

Multiclass classification DL CV Transfer Learning
Notebook
The task is to build classifier using ConvNets to classify images from Simpsons series onto 42 classes.
Implementation includes:

• Data preparation:
  • Label Encoding
• Modeling:
  • Baseline: Custom PyTorch dataset + Torchvision Transforms + Finetuning vgg16_bn
  • Improved model: In progress

Matching subscribers

Multiclass classification Classic ML Tabular Feature Engineering
Notebook
The task is to match customers of telecom company based on their characteristics.
Implementation includes:

• Exploratory Data Analysis
  • Finding & fixing errors in features
  • Memory optimization
  • Checking data integrity
• Modeling:
  • Baseline: TF-IDF + LogisticRegression + GridSearch
  • Final: In progress
• Explaining model predictions: ELI5