Supporting B2Bsales forecasting by machine learning
summary
B2B sales forecasting is framed as a binary classification problem built with CRISP-DM and action design research, trained on 44,448 cases with 22 features (49/51 won/lost, 80/20 train/test) using Orange and R. CRM data prove too subjective so teams re-label in Excel, feature selection shows ~5 top predictors and random forest peaks near nine features; random forest performs best, Naive Bayes second. Coupling model outputs with instance-level explanations (sensitivity-style and set-wise contributions) and participatory workshops measurably improves forecasts and surfaces slip‑causes.
AI-written summary · about
chapters (6)
- 0:00Start
- 5:31Action Design Research Methodology
- 10:15Core of Action Design Research Methodology
- 10:52Data Preparation
- 22:35What-if Analysis
- 24:18Results of Implementation
related talks
37mForecasting using machine learning methods
39mForecasting With Classical and Machine Learning Methods
30mAI in Retail - Advancing Promo Demand Forecasting
22mPredict the future (or don't) with machine learning in applications
31mFrom stockouts to happy customers: Proven solutions for time series forecasting in retail
01h 39mApplying Rigorous Machine Learning Methods in Business Strategy
28mMastering Demand Forecasting: Lessons from Europe's Largest Retailer
56mPredictive Analytics for Practitioners