SAS Viya - developing Code and building Model

SAS Viya for a Customer Churn Prediction Model (a popular business case, especially in telecom and banking) 🚀.

---

💡 Interview-Ready Project Example: Customer Churn Prediction using SAS Viya

"In one of my recent projects, I built a predictive model to identify customers likely to churn, using SAS Viya’s Cloud Analytic Services (CAS). 

This allowed the business to focus retention campaigns on high-risk customers and reduce churn significantly."

---

⚙️ Step-by-Step Approach

---

✅ 1️⃣ Data Preparation in SAS Viya
Loaded the raw data into CAS and cleaned it for missing values, outliers, and inconsistent entries.

cas mySession;
libname casuser cas;

data casuser.churn_data_clean;
    set casuser.customer_data_raw;
    if missing(age) then age = median(age);
    if missing(income) then income = mean(income);
run;

---

✅ 2️⃣ Exploratory Data Analysis (EDA)
Used PROC SGPLOT and Visual Analytics to identify customer churn patterns.

proc sgplot data=casuser.churn_data_clean;
    vbar plan_type / group=churn_event;
run;

---

✅ 3️⃣ Model Building — Gradient Boosting (PROC GRADBOOST)
Built the churn prediction model using SAS’s in-memory machine learning procedure:

proc gradboost data=casuser.churn_data_clean;
    target churn_event / level=nominal;
    input age income tenure / level=interval;
    input region plan_type / level=nominal;
    savestate rstore=casuser.churn_model_gb;
run;

---

✅ 4️⃣ Model Evaluation
Assessed the model's performance with ROC and AUC metrics.

proc cas;
    loadactionset "decisionTree";
    rocTable / table={name="churn_data_clean"}
                rstore="churn_model_gb"
                target="churn_event"
                event="1";
quit;

---

✅ 5️⃣ Model Deployment
Published the trained model to SAS Model Manager for real-time and batch scoring.

proc astore;
    save rstore=casuser.churn_model_gb
         out=casuser.published_churn_model;
run;

The deployed model was integrated with Airtel's customer service systems through REST APIs, allowing real-time churn scoring every time customer data was updated.

---

💡 Business Impact:

"Using this approach in SAS Viya, the business could identify ~78% of high-risk customers before they churned, improving retention through targeted campaigns and saving significant revenue."

---

✅ Why SAS Viya Was the Right Choice:

• Distributed in-memory processing with CAS for speed.

• Integration with Model Manager for model lifecycle governance.

• REST API deployment — no re-coding required for production.

---

🎯 Wrap-Up for Interviews:

"This project showed me the full power of SAS Viya — from data prep, model building, and evaluation, to deployment — all within one scalable ecosystem, which makes it especially efficient for large enterprises like telecom operators or banks."

💡 Interview Answer: Developing Code and Building Models in SAS Viya

"In my recent experience, I’ve used SAS Viya to develop scalable code and build machine learning models in a cloud-native environment. What I like about Viya is that it combines SAS’s traditional strengths in data manipulation and statistics with modern, distributed computing through its CAS (Cloud Analytic Services) engine."

---

🧠 Typical Workflow I Follow in SAS Viya

---

1️⃣ Data Preparation

• I start by loading data into CAS (Cloud Analytic Services) for in-memory distributed processing.

• Clean and transform data using either:

  • SAS Code (PROC SQL, DATA Step).

  • Or open-source languages like Python or R via Viya’s APIs.

Example:

cas mySession;
libname casuser cas;

data casuser.cleaned_data;
   set casuser.raw_data;
   if age >= 18 and status = 'Active';
run;

---

2️⃣ Exploratory Data Analysis (EDA)

• Use PROC CASUTIL, PROC SGPLOT, and SAS Visual Analytics for quick data profiling.

• Helps spot patterns, outliers, and correlations before model building.

---

3️⃣ Model Building

• For modeling, I typically use:

  • PROC FOREST for random forests.

  • PROC GRADBOOST for gradient boosting models.

  • PROC LOGISTIC for classic regression.

  • SAS Viya AutoML or Model Studio for automated model pipelines.

Example: Random Forest Model

proc forest data=casuser.cleaned_data;
   target churn_event / level=nominal;
   input age income tenure / level=interval;
   input region plan_type / level=nominal;
run;

---

4️⃣ Model Evaluation

• I validate model performance using:

  • ROC curves and AUC.

  • Confusion matrices.

  • Lift & Gain charts (available in Model Studio or Visual Analytics).

---

5️⃣ Model Deployment

• Once the model is ready, I deploy it directly through:

  • SAS Model Manager (for versioning, promotion, and governance).

  • Real-time scoring APIs (if integrated with applications).

  • Batch scoring on new datasets.

Example: Model Publishing

proc astore;
   save rstore=casuser.churn_model;
run;

---

💡 Why I Like SAS Viya for Model Development

• Cloud-Native: Supports scale-out computing through Kubernetes and cloud platforms.

• Open Integration: Allows me to write in SAS or use Python / R interchangeably.

• Centralized Governance: Models can be promoted and governed using SAS Model Manager.

• Real-time Deployment: Easy to deploy models via REST APIs, no need to re-code in production.

---

🎯 Wrap-Up for the Interview:

"So in summary, SAS Viya gives me the flexibility to prepare data, build advanced models, evaluate their performance, and deploy them efficiently — all in a distributed, cloud-friendly environment."

---

✅ Pro Tip:
If the interviewer follows up, you can also mention that SAS Viya’s AutoML in Model Studio can automatically select algorithms, tune hyperparameters, and export scoring code.

---

If you want, I can also draft a sample project explanation or code snippet to match a use case like:
👉 fraud detection, customer churn, credit scoring, or sales forecasting.