|
| Inputs: | |
| (i) where No of Arabic text documents |
| (ii) selected document |
| (iii) Where ∀ () ∃ (C is the name of the class) and (j) number of classes |
|
| Output: | Assign to correct class |
|
| Begin | Read all collections of documents in (corpus) |
| ForWHERE I = 1 to n |
| Do pre-processing for document |
| D[I]1 ← Normalization(D[i]) |
| D[I]2 ← stopword removal (D[i]1) IF |
| D[I]3 ← Stemming(D[i]2) IF |
| D[I]4 ← Tokenization(D[i]3) |
| D[I] ← TF-IDF(D[i]3) OR proposed |
| D[I]Train = 80% |
| D[I]Train = 20% |
|
| “Training phase” | W = Input matrix with weights train (TF-IDF/Proposed; document) |
| Weight (W) for document and add label (L) for each document |
| ML ← (W + L) where W is referred to text and L refers to label |
| ENSEMBLE ← (W + L) where W is referred to text and L refers to label |
| DL(Proposed) ← (W + L) where W is referred to text and L refers to label |
|
| “Testing phase” | W = Input matrix with weights train (TF-IDF/Proposed; document) |
| ML ← (W + L) where W is referred to text and L refers to label |
| ENSEMBLE ← (W + L) where W is referred to text and L refers to label |
| DL (proposed) ← (W + L) where W is referred to text and L refers to label |
| End for |
| Push vector value without a corresponding label to classification algorithm then let the algorithm predict L |
| Proposed ensemble transfer Learning ← (text) |
| Class (L) ← proposed ensemble transfer learning (predict) |
|
| End | |
|
