Spaces:
Sleeping
Sleeping
| import os | |
| from shiny import App, ui, render, reactive, module | |
| from shiny import experimental as x | |
| import shinyswatch | |
| import pandas as pd | |
| from modules import lead_ids, classification, ranking, support_texts | |
| # Define lead options | |
| leads_8 = lead_ids.leads_8 | |
| leads_123 = lead_ids.leads_123 | |
| leads_256 = lead_ids.leads_256 | |
| app_ui = ui.page_fluid( | |
| shinyswatch.theme.minty(), | |
| ui.panel_title("Lead Recommender System"), | |
| ui.layout_sidebar( | |
| ui.panel_sidebar( | |
| ui.input_checkbox( | |
| "explanation_select", "Show explanation", True | |
| ), | |
| ui.markdown("**Data**"), | |
| ui.input_select( | |
| "campaign_select", "Select campaign id:", | |
| choices = ["8", "123", "256"], | |
| selected = "8" | |
| ), | |
| ui.input_select( | |
| "lead_select", "Select lead id:", | |
| choices = leads_8 | |
| ), | |
| ui.HTML("<br>"), | |
| ui.markdown("**Classification**"), | |
| ui.input_select( | |
| "model_select", "Select classification model:", | |
| choices = ["BERT", "XGB"], | |
| selected = "BERT" | |
| ), | |
| ui.input_select( | |
| "proba_cutoff_select", "Select minimum relevance probability (%):", | |
| choices = [50, 60, 65, 70, 75, 80, 85, 90, 95], | |
| selected = 80 | |
| ), | |
| ui.HTML("<br>"), | |
| ui.markdown("**Ranking**"), | |
| ui.input_select( | |
| "model_select_rank", "Select ranking model:", | |
| choices = ["Light XGBM 1", "Light XGBM 2", "Light XGBM 3"], | |
| selected = "Light XGBM 3" | |
| ), | |
| ui.input_select( | |
| "rank_cutoff_select", "Select minimum prediction score (%):", | |
| choices = [50, 60, 65, 70, 75, 80, 85, 90, 95], | |
| selected = 80 | |
| ), | |
| width=2.5 | |
| ), | |
| ui.navset_tab( | |
| ui.nav("Classification", | |
| ui.panel_main( | |
| ui.panel_conditional("input.explanation_select", | |
| ui.row( | |
| ui.column( | |
| 12, | |
| x.ui.card( | |
| ui.markdown(support_texts.classification_intro_1), | |
| ui.markdown(support_texts.classification_intro_2), | |
| ui.markdown(support_texts.classification_intro_3), | |
| ), | |
| ), | |
| ), | |
| ), | |
| ui.row( | |
| ui.column( | |
| 12, | |
| x.ui.card( | |
| ui.markdown("**Model classification performance of selected lead**"), | |
| ui.output_data_frame("performance_metrics"), | |
| ), | |
| ), | |
| ), | |
| ui.row( | |
| ui.column( | |
| 12, | |
| x.ui.card( | |
| ui.markdown("**Top 3 relevant employees in selected lead**"), | |
| ui.output_data_frame("predictions_123"), | |
| ), | |
| x.ui.card( | |
| ui.markdown("**All employees in selected lead**"), | |
| ui.output_data_frame("predictions"), | |
| ), | |
| ), | |
| ), | |
| ui.row( | |
| ui.column( | |
| 12, | |
| x.ui.card( | |
| ui.markdown("**Model descriptions and testing performance**"), | |
| ui.markdown(support_texts.models_test_classification_1), | |
| ui.output_data_frame("models_test_classification"), | |
| ui.markdown(support_texts.models_test_classification_2) | |
| ), | |
| ), | |
| ), | |
| ), | |
| ), | |
| ui.nav("Ranking", | |
| ui.panel_main( | |
| ui.panel_conditional("input.explanation_select", | |
| ui.row( | |
| ui.column( | |
| 12, | |
| x.ui.card( | |
| ui.markdown(support_texts.ranking_intro_1), | |
| ui.markdown(support_texts.ranking_intro_2), | |
| ui.markdown(support_texts.ranking_intro_3), | |
| ), | |
| ), | |
| ), | |
| ), | |
| ui.row( | |
| ui.column( | |
| 12, | |
| x.ui.card( | |
| ui.markdown("**Model ranking performance of selected lead**"), | |
| ui.output_data_frame("ndcg_score") | |
| ), | |
| x.ui.card( | |
| ui.markdown("**Top 3 relevant employees in selected lead**"), | |
| ui.output_data_frame("predictions_123_rank"), | |
| ), | |
| ), | |
| ), | |
| ui.row( | |
| ui.column( | |
| 12, | |
| x.ui.card( | |
| ui.markdown("**Ranking of all employees in selected lead**"), | |
| ui.output_data_frame("predictions_rank") | |
| ), | |
| ), | |
| ), | |
| ui.row( | |
| ui.column( | |
| 12, | |
| x.ui.card( | |
| ui.markdown("**Model descriptions and testing performance**"), | |
| ui.markdown("The table below shows the model performance on a larger testing dataset (does not contain the campaigns in this dashboard)"), | |
| ui.output_data_frame("models_test_ranking"), | |
| ui.markdown(support_texts.models_test_ranking_1) | |
| ), | |
| ), | |
| ), | |
| ), | |
| ), | |
| ui.nav("Comparison", | |
| ui.panel_main( | |
| ui.row( | |
| ui.markdown("**Model performance comparison (campaign-level)**"), | |
| ui.column( | |
| 6, | |
| x.ui.card( | |
| ui.markdown("**Classification model performance**"), | |
| ui.output_data_frame("performance_metrics_campaign"), | |
| ui.markdown("<br><br>") | |
| ), | |
| ), | |
| ui.column( | |
| 6, | |
| x.ui.card( | |
| ui.markdown("**Ranking model performance**"), | |
| ui.output_data_frame("campaign_rank") | |
| ), | |
| ), | |
| ), | |
| ui.row( | |
| ui.column( | |
| 12, | |
| x.ui.card( | |
| ui.markdown(support_texts.comparison_1) | |
| ), | |
| ), | |
| ), | |
| ui.row( | |
| ui.column( | |
| 12, | |
| x.ui.card( | |
| ui.markdown(support_texts.ranking_ndcg) | |
| ), | |
| ), | |
| ), | |
| ), | |
| ), | |
| ), | |
| ), | |
| ) | |
| def server(input, output, session): | |
| # Updating lead id selection list | |
| def _(): | |
| if input.campaign_select() == "8": | |
| lead_options = leads_8 | |
| elif input.campaign_select() == "123": | |
| lead_options = leads_123 | |
| elif input.campaign_select() == "256": | |
| lead_options = leads_256 | |
| ui.update_select("lead_select", | |
| label="Select lead id:", | |
| choices=lead_options | |
| ) | |
| ui.update_switch | |
| # Get classification data single lead | |
| def get_lead_predictions(): | |
| try: | |
| df, df_123, df_performance_metrics = classification.classify(CAMPAIGN_ID=int(input.campaign_select()), | |
| LEAD_ID=int(input.lead_select()), | |
| proba_cutoff=int(input.proba_cutoff_select()), | |
| model_type=input.model_select()) | |
| return df, df_123, df_performance_metrics | |
| except TypeError: | |
| pass | |
| except Exception: | |
| ui.notification_show("Data is still loading or something went wrong", duration=3, type="error") | |
| # Get classification data single lead | |
| def get_campaign_predictions(): | |
| try: | |
| df, df_123, df_performance_metrics = classification.classify(CAMPAIGN_ID=int(input.campaign_select()), | |
| LEAD_ID=int(input.lead_select()), | |
| proba_cutoff=int(input.proba_cutoff_select()), | |
| model_type=input.model_select(), | |
| full_campaign=True) | |
| return df, df_123, df_performance_metrics | |
| except TypeError: | |
| pass | |
| except Exception: | |
| ui.notification_show("Data is still loading or something went wrong", duration=3, type="error") | |
| # Get ranking data | |
| def get_lead_ranking(): | |
| try: | |
| df, df_123, ndcg, df_campaign_rank = ranking.rank_single_lead(CAMPAIGN_ID=int(input.campaign_select()), | |
| LEAD_ID=int(input.lead_select()), | |
| rank_cutoff=int(input.rank_cutoff_select()), | |
| ranker=input.model_select_rank()) | |
| return df, df_123, ndcg, df_campaign_rank | |
| except TypeError: | |
| pass | |
| except Exception: | |
| ui.notification_show("Data is still loading or something went wrong", duration=3, type="error") | |
| def predictions_123(): | |
| try: | |
| df, df_123, df_performance_metrics = get_lead_predictions() | |
| return df_123 | |
| except TypeError: | |
| pass | |
| def predictions(): | |
| try: | |
| df, df_123, df_performance_metrics = get_lead_predictions() | |
| return df | |
| except TypeError: | |
| pass | |
| def performance_metrics(): | |
| try: | |
| df, df_123, df_performance_metrics = get_lead_predictions() | |
| return df_performance_metrics | |
| except TypeError: | |
| pass | |
| def performance_metrics_campaign(): | |
| try: | |
| df, df_123, df_performance_metrics = get_campaign_predictions() | |
| return df_performance_metrics | |
| except TypeError: | |
| pass | |
| def predictions_123_rank(): | |
| try: | |
| df, df_123, ndcg, df_campaign_rank = get_lead_ranking() | |
| return df_123 | |
| except TypeError: | |
| pass | |
| def predictions_rank(): | |
| try: | |
| df, df_123, ndcg, df_campaign_rank = get_lead_ranking() | |
| return df | |
| except TypeError: | |
| pass | |
| def ndcg_score(): | |
| try: | |
| df, df_123, ndcg, df_campaign_rank = get_lead_ranking() | |
| return ndcg | |
| except TypeError: | |
| pass | |
| def campaign_rank(): | |
| try: | |
| df, df_123, ndcg, df_campaign_rank = get_lead_ranking() | |
| return df_campaign_rank | |
| except TypeError: | |
| pass | |
| # Model performance tables (test set) | |
| def models_test_classification(): | |
| try: | |
| models_test_classification_data = {'Model': ['BERT', 'XGB'], | |
| 'F1 weighted': [0.84, 0.81], | |
| 'F1': [0.35, 0.35], | |
| 'Accuracy': [0.80, 0.76], | |
| 'Recall': [0.65, 0.78], | |
| 'Precision': [0.24, 0.22]} | |
| df_models_test_classification = pd.DataFrame.from_dict(data=models_test_classification_data, orient='index', columns=models_test_classification_data['Model']) | |
| df_models_test_classification = df_models_test_classification.iloc[1:] | |
| df_models_test_classification.reset_index(inplace=True, names=['Metric']) | |
| return df_models_test_classification | |
| except TypeError: | |
| pass | |
| def models_test_ranking(): | |
| try: | |
| models_test_ranking_data = {'Model': ['Light XGBM 1', 'Light XGBM 2', 'Light XGBM 3'], | |
| 'NDCG@k score': [0.652, 0.2, 0.948], | |
| 'k': [3,3,3]} | |
| df_models_test_ranking = pd.DataFrame.from_dict(data=models_test_ranking_data, orient='index', columns=models_test_ranking_data['Model']) | |
| df_models_test_ranking = df_models_test_ranking.iloc[1:] | |
| df_models_test_ranking.reset_index(inplace=True, names=['Metric']) | |
| return df_models_test_ranking | |
| except TypeError: | |
| pass | |
| app = App(app_ui, server) |