hlz-922/classification-and-model-selection-with-kickstarter

classification-and-model-selection-with-kickstarter

Project Description:

The project aims to classify Kickstarter projects into two categories based on their success or failure, utilizing a variety of features provided in the dataset.

Data Processing:

1. Loading Data: The datasets kickstarter.gz and kickstarter_eval.gz were loaded into pandas DataFrames.
2. Initial Data Inspection: Basic inspection methods such as .info(), .head(), .unique(), and .value_counts()were used to examine the datasets' structure, identifying feature types and potential issues like missing values.
3. Data Cleaning:
   • Dropping unnecessary columns that are unlikely to influence the outcome based on common sense or columns that contain too many nan values, such as 'id', 'photo', 'slug', etc. .drop(cols, axis=1, inplace=True)
   • Filling missing values in textual columns ('name' and 'blurb') with empty strings. df[col].fillna(replacement, inplace=True)
4. Feature Engineering:

   • Extracting categorical data from JSON-like strings in the 'category' column, focusing on general and specific categories.

     • json.loads(x) tidies up the json input
     • pd.json.normalize(df.col.apply(json.loads)) can flatten the JSON strings in one column to a dataframe

   • Handling of textual data and other categorical features to prepare for model input.

     • name/blurb
       • length of string in each row df[col].str.len()
       • word count in each row df[col].str.split().str.len()
       • ave word count in each row df['name'].apply(lambda s: np.mean([len(w) for w in s.split()]))
     • map countries to regions such as asia, europe, etc.

   • One-hot encoding of categorical features and standardization of numerical features to ensure model compatibility.

     from sklearn.preprocessing import OneHotEncoder
     
     ohe = OneHotEncoder(sparse=False, handle_unknown='ignore')
     df_cat = pd.DataFrame(ohe.fit_transform(df_cat))
     new_column_names = ohe.get_feature_names_out(input_features=['region', 'month', 'generic_cat', 'precise_cat'])
     df_cat.columns = new_column_names

Model Selection and Training:

• Logistic Regression: Initially chosen for its simplicity and interpretability. A grid search was conducted to find the optimal hyperparameters, including solver, penalty, and C value.
• Challenges: Convergence issues were encountered during the grid search, likely due to the need for more iterations and/or better feature scaling.