Business Context
A manufacturing plant monitors heavy machinery using IoT sensors. They have a massive dataset of telemetry readings. The engineering team wants to isolate a specific critical component (Main Engine Temperature) and find which other sensors (Vibration, Pressure, RPM) are most strongly correlated with it, to identify leading indicators of overheating. Efficiency is key due to data volume.
Data Preparation
Simulate a high-volume IoT dataset (100,000 observations) with one target variable (Main_Temp) and several potential predictors.
Copied!
| 1 | |
| 2 | DATA mycas.iot_sensors; |
| 3 | call streaminit(999); |
| 4 | DO i=1 to 100000; |
| 5 | Main_Temp = rand('normal', 90, 10); |
| 6 | Vibration = (Main_Temp * 0.5) + rand('normal', 0, 2); |
| 7 | Pressure = rand('uniform', 10, 50); |
| 8 | RPM = 3000 - (Main_Temp * 10) + rand('normal', 0, 50); |
| 9 | OUTPUT; |
| 10 | END; |
| 11 | |
| 12 | RUN; |
| 13 |
Étapes de réalisation
1
Execute correlation analysis targeting specific pairs: 'Main_Temp' vs. all other sensors, avoiding a full N*N matrix computation for performance.
Copied!
| 1 | |
| 2 | PROC CAS; |
| 3 | SIMPLE.correlation TABLE={name='iot_sensors'} inputs={'Vibration', 'Pressure', 'RPM'} pairWithInput={'Main_Temp'}; |
| 4 | |
| 5 | RUN; |
| 6 |
Expected Result
The action returns a focused correlation table listing only the relationships between 'Main_Temp' and the specified input sensors (Vibration, Pressure, RPM). This confirms 'Vibration' has a positive correlation and 'RPM' a negative correlation with temperature, processed efficiently without computing irrelevant sensor-to-sensor correlations.