EXPERIMENT -2
Exploring Data Pane Widgets Using Titanic Dataset
Aim
To study and use different widgets available in the Data Pane by performing operations on
the Titanic dataset.
Steps:
1. Open Orange Data Mining.
2. Open the Data Pane section.
3. Use the File widget to load and view the dataset.
4. Use the Data Table widget to display dataset records in table format.
5. Connect the Paint Data widget and modify or highlight selected data values.
6. Use the Data Info widget to observe dataset details such as rows, columns, features,
and data types.
7. Apply the Rank widget to rank important attributes or values in the dataset.
8. Use the Edit Domain widget to rename or modify attribute properties.
9. Apply the Color widget to assign colors for better visualization of data categories.
10. Use the Column Statistics widget to analyze statistical information like mean,
minimum, maximum, and distribution of columns.
11. Use the Save Data widget to save the modified or processed dataset.
Conclusion:
The widgets helped in loading, viewing, editing, analyzing, coloring, ranking, and saving the
data. This experiment improved understanding of data preprocessing and data analysis
techniques using Data Pane tools.
Lab 7: Recommendation Engine Simulation
Phase 1: Data Preparation & Reduction
1. File Widget: Load your Amazon dataset. Ensure Rating is set as a Numeric variable,
and User ID / Product ID are set as Categorical or Text.
2.
3. Select Rows: Filter the data. To get a clean simulation, you might filter for a specific
range of Product IDs or only include ratings of 4.0 and 5.0 to focus on "positive"
recommendations.
4. Data Sampler: Add this widget to select a random 1% or 0.5% of the data.
Hierarchical clustering calculation time increases exponentially with the number of
rows.
○ Set to: Fixed sample of 1,000–2,000 rows for the smoothest experience.
Phase 2: Scatterplot Visualisation
Distance: Connect the Pivo/Imputed data to the Scatterplot Widget
Phase 3: Output & Recommendation Analysis
Lab 9: Chatbot Intent Classification
1. File Widget (Data Loading)
● Action: Load Bitext_Sample_Customer_Service_Testing_Dataset.csv.
● Column Setup: Double-click the File widget and ensure:
○ utterance: Set to meta or text.
○ intent: Set to target (this is what the model learns to predict).
○ category and tags: Can be set to ignore or meta.
2. Preprocess Text Widget
● Purpose: Cleans the "noise" out of human language.
● Settings:
○ Transformation: Check "Lowercase".
○ Tokenization: Select "Simple" or "Regexp" (breaks sentences into individual
words).
○ Filtering: Select "Stopwords" (removes common words like the, a, is so the
model focuses on meaningful words like refund or track).
3. Logistic Regression Widget
● Action: Connect the Bag of Words to this widget.
● Settings: Leave the default "L2" regularization. This prevents the model from "over-
memorizing" specific sentences so it can handle new, slightly different queries in the
future.
4. Test & Score Widget
● Connections:
○ Connect Bag of Words to Test & Score.
○ Connect Logistic Regression to Test & Score.
● Evaluation: Use Cross-Validation (10 folds).
● Metrics to watch:
○ CA (Classification Accuracy): How often the chatbot is right.
○ F1 Score: A balance between precision and recall (very important for
chatbots).
Activities for Students
● Confusion Matrix: Connect a Confusion Matrix widget to Test & Score. Look for
where the model gets confused. Does it struggle to tell the difference between
cancel_order and track_order?
● Manual Testing: Add a Data Table with a few new sentences you wrote yourself.
Connect it to the Predictions widget along with the trained Logistic Regression model
to see if the chatbot understands your voice.
Outcome
Students will see that "AI understanding" is actually a mathematical process:
Text Vectors (Numbers) Lo
Lab 10: Sentiment Analysis of Reviews
1. Data Loading (The Entry Point)
Every process starts with getting data into the system.
● File Widget: Used to browse and load your .csv, .xlsx, or .tab files.
● Data Properties: You must define which column is your
● Target (what you want to predict) and which are
● Features (the data used to make the prediction).
2. Data Preprocessing (Cleaning)
Raw data is rarely perfect. This stage prepares the data for mathematical models.
● Select Columns: To remove unnecessary data or change the target variable.
● Impute: To fill in missing values (NaN) with averages or zeros.
● Preprocess Text: (For NLP) To lowercase text, remove punctuation, and filter out
"stopwords" (like and, the, is).
3. Data Visualization (Exploration)
Before modeling, you look for patterns or outliers.
● Scatter Plot: To see clusters or correlations between two variables.
● Distributions: To see the spread of your data (e.g., how many "Positive" vs
"Negative" reviews you have).
4. Modeling (The AI Engine)
This is where the machine "learns." You connect your processed data to a learning
algorithm.
● Classification: Using SVM, or Naive Bayes to put data into categories.
● Clustering: Using k-Means or Hierarchical Clustering to find hidden groups.
● Regression: Using Linear Regression to predict a specific number.
Output: