Multidataset

lib/wekafiles/packages/new_package_check.txt

@@ -1 +1 @@
-1598950809434
+1599209053791

ubumlaas/experiments/algorithm/__init__.py

@@ -22,6 +22,9 @@
 from time import time
 import json
 import os
+from ubumlaas.util import string_is_array
+
+from copy import deepcopy
 
 import weka.core.jvm as jvm
 from ubumlaas.util import get_dataframe_from_file
@@ -49,55 +52,88 @@ def task_skeleton(experiment, current_user):
     # Diference sklearn executor and weka executor
     # Get algorithm type
     data = experiment["data"]
+    data = string_is_array(data)
+    if not isinstance(data,list):
+        data = [data]
+    v.app.logger.debug("%d - %d - Data list know is a list - %s", current_user['id'], experiment['id'], data)
+    # From here data will be a list
     if exp_manager.is_multi:
         rep = exp[0]["exp_config"]["repetition"]
     else:
         rep = exp["exp_config"]["repetition"]
-    res_global = []
-    state_global=[]
-    for i in range(rep):
-        if exp_manager.is_multi:
-            res=[]
-            state=[]
-            #Seed for no seed experiments, multi experiments should execute with the same data
-            rand = random.randint(1,1000000000)        
-            for i in range(len(exp)):
-                r,s = execute_model(exp[i]["alg"]["lib"],data,exp[i],current_user,rand)
-                if s == 1:
-                    res.append(json.loads(r))
+    v.app.logger.debug("%d - %d - Repetitions: %d", current_user['id'], experiment['id'], rep)
+    res_global_data = []
+    state_global_data=[]
+    for j in range(len(data)):
+        res_global = []
+        state_global=[]
+        for _i in range(rep):
+            if exp_manager.is_multi:
+                res=[]
+                state=[]
+                #Seed for no seed experiments, multi experiments should execute with the same data
+                rand = random.randint(1,1000000000)        
+                for i in range(len(exp)):
+                    aux_exp = deepcopy(exp[i])
+                    if len(data)>1:
+                        aux_exp["exp_config"]["target"]=aux_exp["exp_config"]["target"][j]
+                        aux_exp["exp_config"]["columns"]=aux_exp["exp_config"]["columns"][j]
+                    r,s = execute_model(exp[i]["alg"]["lib"],data[j],aux_exp,current_user,rand)
+                    if s == 1:
+                        res.append(json.loads(r))
+                    else:
+                        res.append(r)
+                    state.append(s)
+                if 2 in state:
+                    state = 2
                 else:
-                    res.append(r)
-                state.append(s)
-            if 2 in state:
-                state = 2
+                    state = 1
+                res_global.append(res)
             else:
-                state = 1
-            res_global.append(res)
+                aux_exp = deepcopy(exp)
+                if len(data)>1:
+                        aux_exp["exp_config"]["target"]=aux_exp["exp_config"]["target"][j]
+                        aux_exp["exp_config"]["columns"]=aux_exp["exp_config"]["columns"][j]
+                res,state=execute_model(exp["alg"]["lib"],data[j],aux_exp,current_user)               
+                if state == 1:
+                    res_global.append(json.loads(res))
+                else:
+                    res_global.append(res)
+
+            state_global.append(state)
+        if 2 in state_global:
+            state_global = 2
         else:
-            res,state=execute_model(exp["alg"]["lib"],data,exp,current_user)
-            res_global.append(json.loads(res))
-
-        state_global.append(state)
-    if rep == 1:
-        res_global = res_global[0]
+            state_global = 1
+        if rep == 1:
+            res_global = res_global[0]
+
+        res_global_data.append(res_global)
+        state_global_data.append(state_global)
 
     #Calculate result means
-    res = res_global
-    if rep>1 and 2 not in state_global:
-        state=1        
-        if exp_manager.is_multi:
-            res_mean = []
-            for i in range(len(res[0])):
-                res_aux = []
-                for j in range(len(res)):
-                    res_aux.append(res[j][i])
-                r_mean = calc_res_mean(res_aux, rep)
-                res_mean.append(r_mean)            
+    res = res_global_data
+    res_data=[]
+    for k in range(len(data)):
+        if rep>1 and 2 not in state_global_data:
+            state=1        
+            if exp_manager.is_multi:
+                res_mean = []
+                for i in range(len(res[k][0])):
+                    res_aux = []
+                    for j in range(len(res[k])):
+                        res_aux.append(res[k][j][i])
+                    r_mean = calc_res_mean(res_aux, rep)
+                    res_mean.append(r_mean)            
+            else:
+                res_mean=calc_res_mean(res[k], rep)
         else:
-            res_mean=calc_res_mean(res, rep)
-
-        res=res_mean
-    elif 2 not in state_global:
+            res_mean=res[k]
+        res_data.append(res_mean)
+    res=res_data
+    if len(data)==1:
+        res = res[0]
+    elif 2 not in state_global_data:
         state = 1
     else:
         state = 2
@@ -121,18 +157,21 @@ def calc_res_mean(res, rep):
     Returns:
         dict: Results mean
     """
+    print(res)
     res_mean={}
     for i in res[0].keys():
-        aux=[]
-        for j in range(rep):
-            aux.append(res[j][i])
-        if isinstance(res[0][i],list):
-            res_mean[i]= np.array(aux).mean(0).tolist()
-        else:
-            res_mean[i]=np.array(aux).mean().tolist()
+        if i != "ROC": # ROC array incorrect over mean, represent points in space, not value. Use AUC to know the mean.
+            aux=[]
+            for j in range(rep):
+                aux.append(res[j][i][0])
+            if isinstance(res[0][i],list):
+                res_mean[i]= np.array(aux).mean(0).tolist()
+            else:
+                res_mean[i] = np.array(aux).mean().tolist()
+            res_mean[i] = [res_mean[i]]
     return res_mean
 
-def execute_model(alg_lib,data, exp, current_user,seed_multi=None):
+def execute_model(alg_lib, data, exp, current_user,seed_multi=None):
     type_app = alg_lib
     execution_lib = None
     try:
@@ -161,7 +200,13 @@ def execute_model(alg_lib,data, exp, current_user,seed_multi=None):
         y_score_list = []
         y_test_list = []
         X_test_list = []
-        if exp_config.get("mode") == "split" and exp_config["train_partition"] < 100:
+
+        if execution_lib.algorithm_type == "Clustering":
+            y_pred, y_score = execution_lib.predict(model, X)
+            y_pred_list.append(y_pred)
+            y_score_list.append(y_score)
+
+        elif exp_config.get("mode") == "split" and exp_config["train_partition"] < 100:
             X_train, X_test, y_train, y_test = execution_lib\
                 .generate_train_test_split(X, y, exp_config["train_partition"],random_state=seed_multi)
             model = execution_lib.create_model()
@@ -185,10 +230,7 @@ def execute_model(alg_lib,data, exp, current_user,seed_multi=None):
                 y_test_list.append(y_test)
                 X_test_list.append(X_test)
 
-        elif execution_lib.algorithm_type == "Clustering":
-            y_pred, y_score = execution_lib.predict(model, X)
-            y_pred_list.append(y_pred)
-            y_score_list.append(y_score)
+
 
         score = {}
         if exp_config["mode"] == "cross" or exp_config["train_partition"] < 100 or execution_lib.algorithm_type == "Clustering":

ubumlaas/experiments/forms.py

@@ -34,11 +34,13 @@ class ExperimentForm(FlaskForm):
     filter_name = SelectField("Select Filter",
                               choices=[("", "---")])
 
-    def __init__(self, id=None):
+    def __init__(self, id=None, iddata=None):
         super(ExperimentForm, self).__init__()
         if id is not None:
             self.alg_name.id="alg_name_"+str(id)
             self.filter_name.id="filter_name_"+str(id)
+        if iddata is not None:
+            self.data.id="data_"+str(iddata)
 
     def alg_list(self, alg_typ):
         """Generate a list of supported algorithms by type.

ubumlaas/experiments/views/dataset.py

@@ -1,5 +1,5 @@
 from flask import \
-    (render_template)
+    (render_template, request)
 from ubumlaas.experiments.forms import \
     (ExperimentForm, DatasetForm)
 from flask_login import (current_user, login_required)
@@ -24,6 +24,16 @@ def update_dataset_list():
     form_e.dataset_list() #log inside dataset_list
     return render_template("blocks/select_dataset.html", form_e=form_e)
 
+@login_required
+@views.experiments.route("/new_dataset_maker", methods=["POST"])
+def new_dataset_maker():
+
+    iddex = request.form.get("idex", 0)
+    form_e = ExperimentForm(0, iddex)
+    form_e.dataset_list()
+    form_d = DatasetForm()
+    return  render_template("blocks/dataset_maker.html", form_e=form_e, iddex=iddex, form_d=form_d);
+
 
 @login_required
 @views.experiments.route("/new_experiment/new_dataset", methods=["POST"])

ubumlaas/experiments/views/experiment.py

@@ -29,7 +29,7 @@ def new_experiment():
         str -- HTTP response with rendered page.
     """
 
-    form_e = ExperimentForm(0)
+    form_e = ExperimentForm(0, 0)
     form_e.dataset_list()
 
     form_d = DatasetForm()
@@ -45,7 +45,7 @@ def new_experiment():
 
     return render_template("experiment_form.html", form_e=form_e,
                            form_d=form_d, form_p=form_p,
-                           title="New experiment", experiment=experiment, idex=0)
+                           title="New experiment", experiment=experiment, idex=0, iddex=0)
 
 
 @login_required
@@ -74,12 +74,15 @@ def launch_experiment():
         if filter_name is not None:
             filter_name = "["+filter_name+"]"
             filter_config = "["+filter_config+"]"
+    data = request.form.get("data")
+    if "," in data:
+        data = "["+data+"]"
 
     exp = Experiment(user.id, alg_name,
                      unquote(alg_config),
                      json.dumps(exp_config),
                      filter_name, filter_config,
-                     request.form.get("data"),
+                     data,
                      None, time.time(), None, 0)
     v.app.logger.info("%d - Create experiment", current_user.id)    
     v.db.session.add(exp)
@@ -144,13 +147,14 @@ def change_column_list():
     Returns:
         str -- HTTP response with JSON
     """
-    form_e = ExperimentForm()
+    iddex = request.form.get("iddex",0)
+    form_e = ExperimentForm(iddex)
     filename = form_e.data.data
     upload_folder = "ubumlaas/datasets/"+current_user.username+"/"
     df, target_columns = get_dataframe_from_file(upload_folder, filename, target_column=True)
-    to_return = {"html": render_template("blocks/show_columns.html", data=df),
+    to_return = {"html": render_template("blocks/show_columns.html", data=df, iddex=iddex),
                  "html2": render_template("blocks/show_columns_reduced.html",
-                                          data=df.columns),
+                                          data=df.columns, iddex=iddex),
                  "df": generate_df_html(df),
                  "config": target_columns}
     v.app.logger.info("%d - Get dataset information - %s%s%s", current_user.id, upload_folder,"/",filename)

ubumlaas/static/css/bootstrap-append.css

@@ -208,4 +208,14 @@ footer .main-footer {
     -ms-transition: all 250ms ease-in-out 0s;
     -o-transition: all 250ms ease-in-out 0s;
     transition: all 250ms ease-in-out 0s;
-  }
+  }
+
+.custom-select[multiple],
+.custom-select[size]:not([size="1"])
+{
+    height: auto;
+    -webkit-appearance: menulist-textfield;
+    background-image: none;
+    /* Fix for excess right padding */
+    padding-right: 0.75rem;
+}

ubumlaas/static/css/custom.css

@@ -106,17 +106,17 @@ option.list-group-item{
     padding: 0.25rem 1rem !important;
 }
 
-#sel[multiple] option.list-group-item-success:checked{
+.sel[multiple] option.list-group-item-success:checked{
     background-color: var(--success) !important;
     color: #ffffff !important;
 }
 
-#sel[multiple] option.list-group-item-primary:checked{
+.sel[multiple] option.list-group-item-primary:checked{
     background-color: var(--primary) !important;
     color: #ffffff !important;
 }
 
-#sel[multiple]  option.list-group-item-secondary:checked{
+.sel[multiple]  option.list-group-item-secondary:checked{
     background-color: var(--secondary) !important;
     color: #ffffff !important;
 }
@@ -125,14 +125,13 @@ tr.align-middle td{
     vertical-align: middle !important;
 }
 
-#makers{
+#makers, #data-makers{
     position: relative !important;
     display: flex;
     align-items: stretch;
-
 }
 
-.maker{
+.maker, .data-maker{
     position: static;
     height: 100%;
     width: 100%;