Architecture Design

Data Preparation

Model Serving

Model Training

Requirement Gathering

Pay Range on Glassdoor

Predict Trip ETA on Uber

ML System Design


<p><strong>Data Sources</strong></p>



<p>Assume that you have the following data sources available to you.</p>



<p><strong>Intersection</strong></p>



<figure class="wp-block-table"><table class="has-fixed-layout"><thead><tr><th>Variable</th><th>Data Type</th><th>Example</th></tr></thead><tbody><tr><td>Location ID</td><td>UUID</td><td>1XAA-BDDS-XX41</td></tr><tr><td>Signal</td><td>Boolean</td><td>Yes</td></tr><tr><td>Number of Lanes</td><td>Numeric</td><td>4</td></tr><tr><td>Pedestrian Crossing</td><td>Boolean</td><td>Yes</td></tr></tbody></table></figure>



<p><strong>Road Segment</strong></p>



<figure class="wp-block-table"><table class="has-fixed-layout"><thead><tr><th>Variable</th><th>Data Type</th><th>Example</th></tr></thead><tbody><tr><td>Location ID</td><td>UUID</td><td>3430-33XD-4443</td></tr><tr><td>Intersection A ID</td><td>UUID</td><td>1XAA-BDDS-XX41</td></tr><tr><td>Intersection B ID</td><td>UUID</td><td>1XAA-4FDK-ZX41</td></tr><tr><td>Number of Lands</td><td>Numeric</td><td>Yes</td></tr><tr><td>One Way</td><td>Boolean</td><td>No</td></tr><tr><td>Segment Type</td><td>Categorical</td><td>(Highway, Road, Dirt)</td></tr><tr><td>Length (Feet)</td><td>Numeric</td><td>50</td></tr><tr><td>Speed Limit (MPH)</td><td>Numeric</td><td>70</td></tr></tbody></table></figure>



<p><strong>Driver Information</strong></p>



<figure class="wp-block-table"><table class="has-fixed-layout"><thead><tr><th>Variable</th><th>Data Type</th><th>Example</th></tr></thead><tbody><tr><td>Driver ID</td><td>UUID</td><td>34GA</td></tr><tr><td>Vehicle Type</td><td>Categorical</td><td>(SUV, Sedan)</td></tr><tr><td>Ratings</td><td>Numeric</td><td>4.9</td></tr><tr><td>Age</td><td>Numeric</td><td>68</td></tr></tbody></table></figure>



<p><strong>Historical Trip</strong></p>



<figure class="wp-block-table"><table class="has-fixed-layout"><thead><tr><th>Variable</th><th>Data Type</th><th>Example</th></tr></thead><tbody><tr><td>Trip ID</td><td>UUID</td><td>566FJA-X4315</td></tr><tr><td>Start ID</td><td>UUID</td><td>433D-55XD-7743</td></tr><tr><td>End ID</td><td>UUID</td><td>3430-33XD-4443</td></tr><tr><td>Start Time</td><td>Timestamp</td><td>July 28, 2023 15:33:25.15</td></tr><tr><td>End Time</td><td>Timestamp</td><td>July 28, 2023 15:53:25.15</td></tr><tr><td>Distance (Mile)</td><td>Numeric</td><td>5.3</td></tr><tr><td>Time (Minutes)</td><td>Numeric</td><td>20</td></tr></tbody></table></figure>



<p><strong>Weather Data</strong></p>



<figure class="wp-block-table"><table class="has-fixed-layout"><thead><tr><th>Variable</th><th>Data Type</th><th>Example</th></tr></thead><tbody><tr><td>Location ID</td><td>UUID</td><td>3430-33XD-4443</td></tr><tr><td>Timestamp</td><td>Timestamp</td><td>July 28, 2023 15:53:25.15</td></tr><tr><td>Condition</td><td>Categorical</td><td>(Sunny, Rain, Foggy)</td></tr></tbody></table></figure>



<p><strong>Feature Engineering</strong></p>



<p><strong>Intersection ETA Model Features &#8211;</strong>&nbsp;Features that are useful in predicting the average waiting time at an intersection</p>



<figure class="wp-block-table"><table class="has-fixed-layout"><thead><tr><th>Features</th><th>Data Type</th><th>Data Source</th><th>Feature Engineering</th></tr></thead><tbody><tr><td>Signal</td><td>Boolean</td><td>Intersection</td><td>One Hot Encoding</td></tr><tr><td>Number of Lanes</td><td>Numeric</td><td>Intersection</td><td>Mean Imputation</td></tr><tr><td>Pedestrian Crossing</td><td>Boolean</td><td>Intersection</td><td>One Hot Encoding</td></tr><tr><td>Vehicle Type</td><td>Categorical</td><td>Driver</td><td>One Hot Encoding</td></tr><tr><td>Ratings</td><td>Numeric</td><td>Driver</td><td>Mean Imputation</td></tr><tr><td>Age</td><td>Numeric</td><td>Driver</td><td>Mean Imputation</td></tr><tr><td>Start Time</td><td>Time</td><td>Historical Trip</td><td>Time Decomposition (year, month, day, hour,…)</td></tr><tr><td>Condition</td><td>Categorical</td><td>Weather</td><td>One Hot Encoding</td></tr></tbody></table></figure>



<p><strong>Road Segment ETA Model Features &#8211;</strong>&nbsp;Features that are useful in predicting the average traveling time on a road segment from one intersection to the next.</p>



<figure class="wp-block-table"><table class="has-fixed-layout"><thead><tr><th>Features</th><th>Data Type</th><th>Data Source</th><th>Feature Engineering</th></tr></thead><tbody><tr><td>Number of Lands</td><td>Numeric</td><td>Road Segment</td><td>Mean Imputation</td></tr><tr><td>One Way</td><td>Boolean</td><td>Road Segment</td><td>One-Hot Encoding</td></tr><tr><td>Segment Type</td><td>Categorical</td><td>Road Segment</td><td>One-Hot Encoding</td></tr><tr><td>Length (Feet)</td><td>Numeric</td><td>Road Segment</td><td>Mean Imputation</td></tr><tr><td>Speed Limit (MPH)</td><td>Numeric</td><td>Road Segment</td><td>Mean Imputation</td></tr><tr><td>Vehicle Type</td><td>Categorical</td><td>Driver</td><td>One Hot Encoding</td></tr><tr><td>Ratings</td><td>Numeric</td><td>Driver</td><td>Mean Imputation</td></tr><tr><td>Age</td><td>Numeric</td><td>Driver</td><td>Mean Imputation</td></tr><tr><td>Start Time</td><td>Time</td><td>Historical Trip</td><td>Time Decomposition (year, month, day, hour,…)</td></tr><tr><td>Condition</td><td>Categorical</td><td>Weather</td><td>One Hot Encoding</td></tr></tbody></table></figure>



<p><strong>Feature Selection</strong></p>



<p>You can consider filter, wrapper or model-based feature selection that utilizes L1, L2, or Elastic-Net regularization. Dimensionality reduction using PCA could be considered too.</p>


Variable	Data Type	Example
Location ID	UUID	1XAA-BDDS-XX41
Signal	Boolean	Yes
Number of Lanes	Numeric	4
Pedestrian Cro...

Data Preparation

Join now to continue learning