Turi Create is a machine-learning library with high-level interfaces for preparing data, training models, and making predictions. Its SFrame stores tabular data. section 3.5
Example workflows show how the library supports different models:
- Train a linear regression model to predict Hyderabad house prices, then use it to estimate a price from area and bedroom count. section 3.5
- Expand a feature into powers for polynomial regression, split data into training and test sets, and set L1 or L2 penalties. section 4.6
- Turn IMDB review text into word-count features and train a logistic classifier to predict positive or negative sentiment. section 6.4
A house-price prediction
A model trained with area and bedroom count predicts a price of 2,594,841 for a house with area 1,000 and three bedrooms. section 3.5
Table 3.3 The first five rows and seven columns of the Hyderabad housing prices dataset (view table figure)
| Price |
Area |
No. of Bedrooms |
Resale |
MaintenanceStaff |
Gymnasium |
SwimmingPool |
| 30000000 |
3340 |
4 |
0 |
1 |
1 |
1 |
| 7888000 |
1045 |
2 |
0 |
0 |
1 |
1 |
| 4866000 |
1179 |
2 |
0 |
0 |
1 |
1 |
| 8358000 |
1675 |
3 |
0 |
0 |
0 |
0 |
| 6845000 |
1670 |
3 |
0 |
1 |
1 |
1 |
The sample rows show housing features such as area and number of bedrooms alongside prices. section 3.5
