1 Introduction
2 Literature review
3 Problem definition and mathematical formulation
Inter-region request | |||
---|---|---|---|
Yes | No | ||
Intra-region request | Yes | Scheduling \(+\) VRPMB \(+\) multi-depot PDP | Multi-depot PDP |
No | Scheduling \(+\) VRPMB |
-
U set of regions = \(\{1, 2\}\)
-
\(N^{u}\) set of inter-region requests starting in region \(u = \{1,\ldots ,n\}\)
-
\(R^{u}\) set of intra-region requests served in region \(u = \{1,\ldots ,n\}\)
-
\(P^{u}\) depots in region u
-
\(V^{u}\) set of nodes to visit in each region u, including depots
-
\(K_{j}^{u}\) set of LDTs in depot j in region u
-
\(K^{u}\) \(\cup _{j \in L^{u}}K_{j}^{u}\)
-
\(L_{j}^{u}\) set of HDVs in depot j from region u
-
T set of days
-
\(c_{ii'}^{u}\) travel cost between points i and \(i'\) in region u
-
\(d_{jj'}^{u}\) travel cost between depots \(j\in P^{u}\) and \(j' \in P^{{\bar{u}}}\)
-
\(\varGamma \) fixed cost for using a LDT
-
\(D_{i}\) demand of request i
-
\(q_{i}^{u}\) quantity load on customer i in region \(u {\left\{ \begin{array}{ll} D_{i}, \text {if it is a pick up point}\\ -\,D_{i}, \text { if it is a delivery point} \end{array}\right. }\)
-
\(E_{i}\) earliest pickup day for request i
-
\(A_{i}\) latest delivery day of request i
-
\(Q_\mathrm{SH}\) load limitation for LDTs
-
\(Q_\mathrm{LH}\) load limitation for HDVs
-
\(X_{ii'k}^{ut}\) \({\left\{ \begin{array}{ll} 1, \hbox { arc }i-i'\hbox { in region }u\hbox { covered by vehicle }k \in K^{u} \hbox { on day }t, i,i' \in V^{u} \\ 0, \text { otherwise} \end{array}\right. }\)
-
\(Y_{ijj'l}^{ut}\) \({\left\{ \begin{array}{ll} 1,\hbox { request }i \hbox { transported in HDV }l \hbox { on period }t, j\in P^{u}, j'\in P^{{\bar{u}}}\\ 0, \text { otherwise} \end{array}\right. }\)
-
\(Z_{k}^{ut}\) \({\left\{ \begin{array}{ll} 1,\hbox { if LDT }k \in K^{u}\hbox { is used on day }t\\ 0, \text { otherwise} \end{array}\right. }\)
-
\(O_{jj'l}^{ut}\) \({\left\{ \begin{array}{ll} 1, \hbox { HDV }l\hbox { from }j \in P^{u} \hbox { to }j' \in P^{{\bar{u}}} \hbox { used on period }t\\ 0, \text {otherwise}\end{array}\right. }\)
-
\(Q_{ik}^{ut}\) \(\hbox { Quantity loaded after visiting node } i \in V^{u}\hbox { in vehicle }k \in K^{u}\hbox { on period } t\)
-
\(S_{ik}^{ut}\) \(\hbox { Time of service in node }i \in V^{u}\hbox { by vehicle }k\hbox { on period }t\)
4 Methods
-
Acceptance criteria 1 New solution \(x'\) is accepted as new incumbent solution with probability \(e^{(f(x) - f(x'))/T}\). Parameter T controls this probability of acceptance along the search. It is initialized with value \(T_{0} = 0.05\times f(x)\times \log (0.5)\) and updated by a decreasing factor of \(T\_\mathrm{cooling}\) after each iteration.
-
Acceptance criteria 2 The new solution \(x'\) is also accepted as new incumbent solution in case that \(f(x') < f(x) + 0.5\). This helps to maintain a path toward a local optima when a new incumbent solution is accepted with acceptance criteria 1.
4.1 ALNS operators
4.1.1 Operators for the scheduling problem
-
Removal random inter Inter-region requests are randomly removed.
-
Removal worst inter This operator selects those inter-region requests that generate a higher cost on the solution. The cost of an inter-region request in the solution is calculated as the proportional cost of this request in the HDV and the cost of its customers in their respective short-haul routes, \(C_{r} = C^\mathrm{LH}_{r} + C^\mathrm{SH}_{r,p} + C^\mathrm{SH}_{r,d}\), where indices p and d stand for pickup an delivery customer, respectively. The cost proportion of request r in its HDV is calculated as the share of r in the total quantity transported by the vehicle, \(C^\mathrm{LH}_{r} = \mathrm{LHCost} \times (r_\mathrm{demand}/LH_\mathrm{quantityTransported})\). The cost of a customer node of r in its short-haul route is the difference in distance between the current route and the same route without this customer. In each removal step, a list of the requests with higher cost is generated and sorted. To randomize the search, one of the inter-region requests is selected following a probability rule from the top of the list.
-
Removal related inter This removal strategy seeks to remove requests that show a high similarity. The idea is that removing requests considered similar would lead to a highest degree of flexibility and interchange in the later repair step. Similarity between requests r and \(r'\) is measured by a relatedness score \(R_{rr'}\) calculated as a combination of two factors: geographical distance and due date distance. The first one is the distance between the pickup and delivery customers of the requests, \(D_{rr'} = d_{r_{p}r'_{p}} + d_{r_{d}r'_{d}}\). The due day distance is the difference in the due dates for servicing each request, \(P_{rr'} = |E_{r} - E_{r'}| + |A_{r}-A_{r'}|\). Both distance measures are normalized by dividing the obtained raw value by the maximum possible distance, in terms of geographical locations and due dates. Finally, \(R_{rr'} = D_{rr'} + P_{rr'}\). The operator starts by randomly selecting a request to remove from the route. In each step, the relatedness value is computed between the last removed request and each request in the solution. The request with highest score is then removed. The process continues until the desired number of removed requests is reached.
-
Removal HDV inter In this operator, entire HDVs and their transported requests are removed from the solution. To avoid an excessive rate of destruction in the solution, the destruction rate is selected randomly between 10 and 20% of the total number of HDVs. Inside the operator, two removal strategies are defined. In the first one, a vehicle is randomly selected and removed. The second strategy removes vehicles incurring higher costs. The costs for a vehicle l are measured as the average costs of all its transported requests, \(C_{l} = \sum _{r\in l} C^\mathrm{LH}_{r}/|l|\), where |l| is the number of requests transported by l. In each step, one of these strategies is selected with probability 50%.
-
Removal proximity inter This operator prioritizes the removal of inter-region requests whose customers are close to other depots in the corresponding region. The goal is to enhance the reallocation of requests to HDVs traveling to depots where the customers of the request might find a good routing position.
-
Repair greedy insertion inter In this operator, the reinsertion of removed inter-region requests follows a greedy strategy. For each request in the reinsertion pool, the insertion position with the least cost is calculated. The cost of insertion of a request to HDV l is calculated using the same equation as in the removal worst operators. That is, the shared cost of the request if transported by vehicle l plus the cost derived from adding the customers to an existing route or from creating a new route. When a new route is created, a fixed cost per vehicle is added to the distance cost. To decrease the deterministic outcome of the method, we create a list with all requests best insertion positions and randomly select one of the top. The process is repeated until all requests have been assigned to an HDV.
-
Repair regret inter The k-regret insertion operators aim at selecting those requests with a bigger difference between the cost of the best insertion position and that of the k-best one. The idea behind it is that not inserting a request with a big cost difference might lead to bad insertion afterward. The costs are calculated in the same way as in the repair greedy insertion operator. Also a list of all requests along with their regret value is created and a request is randomly chosen from the top of the list. Our ALNS algorithm makes use of 2-regret and 3-regret operators.
4.1.2 Operators for the multi-depot PDP
-
Removal random intra In each step, a randomly chosen intra-region request is removed, until we reach the desired quantity of requests to remove.
-
Removal worst intra This operator works in the same way as removal random inter. In this case, the cost of an intra-region request in the solution is calculated as the difference in distance cost in the routing with and without the request on it. Since the intra-region request has both customers in the routing plan of the same depot, cost \(C_{r}\) is the cost that both nodes are generating in their corresponding routes, \(C_{r} = C_{r,p} + C_{r,d}\).
-
Removal related intra Again we use the same strategy as in the scheduling operator counterpart, removal related inter. The way of measuring the relatedness score between two requests is also identical, only that now we measure the scores of nodes situated in the same region.
-
Repair greedy insertion intra This operator works similar to repair greedy insertion inter, but the costs are calculated as in removal worst intra. When inserting an intra-region request, all insertion positions in the routing plan that satisfy the due dates of the request, are evaluated. Afterward, the best combination of positions for the pickup and delivery nodes are selected, such that it satisfies the precedence constraints, if necessary.
-
Repair regret intra Again we use a similar strategy to repair regret inter. Insertion costs and feasible positions are calculated as described in repair greedy insertion intra. 2-regret and 3-regret operators are used.
4.1.3 Operators for the VRPMB \(+\) PDP
-
Removal random sort-haul This operator removes customers at random.
-
Removal worst short-haul Similar to the previously described operators. In this case, the cost that a customer generates in a solution is calculated as the difference in distance cost in the route where it belongs, when comparing with the same route without the customer. If a customer is served in position i of a route, then \(C_{i} = d(i-1,i) + d(i,i+1) - d(i-1,i+1)\), where d(i, j) is the distance between customers in positions i and j in the route. The customers with higher cost are removed from the solution following a randomized strategy as in the previous strategies.
-
Removal related short-haul Same concept as removal related inter, but here the relatedness score between customers s and \(s'\) is calculated as the sum of geographical distance \(d_{ss'}\), customer due dates distance, and difference in visit time. The due date distance is calculated by the feasible days to serve a customer, in the same way as in removal related inter. As previously mentioned, feasible service days for customers from inter-region requests are determined by the due dates of the request and the day when the long-haul is transported. In case of customers from intra-region requests, the feasible service days are determined by the request’s due date and the day of service of the other customer of the request. That is, a delivery customer cannot be served on a day before the pickup occurs, and vice versa. The visit time distance is measured as \(|b_{s} - b_{s'}|\), where \(b_{s}\) is the visit time of customer s. The customer to remove in each round is selected identically as in the other removal related algorithms.
-
Removal route short-haul This operator removes entire routes from the depot routing plan. As in removal long-haul vehicle inter, the removal rate is adjusted to 10–20% to avoid destroying too much. Also two different processes for removal can be used. One process removes a route randomly. The second process removes those routes with higher cost. The cost of route v is calculated as the average distance cost per customer in the route, \(C_{v} = \sum _{i\in v} d(i,i+1)/|v|\), where |v| is the number of customers in route v. In each removal round, each process is selected with a probability of 50%.
-
Repair greedy insertion short-haul The customers in the customer pool are reinserted in their cheapest feasible positions. The insertion cost is calculated in the same way as in removal worst short-haul. A fixed cost per vehicle is added to the distance cost when a new route must be created. A randomized selection of the best possible insertion positions is done in each round.
-
Repair regret short-haul This operator works exactly in the same manner as the previously described regret operators. The cost structures are the same as in removal worst short-haul. We use 2-regret and 3-regret as insertion regret operators.
4.2 Local search
-
Swap The swap strategy tries to interchange two elements from their current positions. In each iteration, a first element is randomly selected and all possible interchanges with other elements are evaluated until a feasible swap improving the objective value is found.
-
Relocate This operator searches for relocations of randomly chosen elements to other places in the solution, providing that the change leaves the solution feasible and the move leads to an improvement in the objective function.
4.3 Obtaining an initial solution
5 Computational experiments
5.1 Algorithm performance for considered subproblems
5.1.1 Results for the VRPTW
Benchmark | ALNS for VRPMB \(+\) PDP | |||
---|---|---|---|---|
nVeh | z | nVeh | Gap (%) | |
C1 | 10 | 828.38 | 10 | 0 |
C2 | 3 | 589.85 | 3 | 0 |
R1 | 11.91 | 1210.33 | 11.91 | 1.54 |
R2 | 2.72 | 951.03 | 2.72 | 1.79 |
RC1 | 11.5 | 1384.16 | 11.5 | 1.33 |
RC2 | 3.25 | 1119.24 | 3.25 | 1.58 |
5.1.2 Results for the LRP
-
The first region contains only the hub as a depot, while the second one contains all possible depot locations for the LRP instance.
-
All LRP requests are transformed into inter-region requests, all of them with the pickup node located in the LRP hub and the delivery node being the customer location in the LRP instance.
-
Only one long-haul vehicle per lane is allowed. The capacity of the long-haul vehicle is the capacity of the potential depot in the region, where it arrives. The cost of the long-haul trip is the opening cost of this depot.
H (%) | Scheduling \(+\) VRPMB (%) | |
---|---|---|
Barreto | 0.21 | 0.42 |
Prins | 0.79 | 0.42 |
Tuzun | 0.57 | 0.74 |
5.2 Experimental data
Instance | #Inter-region requests | #Intra-region requests | #Depots | #Days | Vehicle capacities |
---|---|---|---|---|---|
T01 | 50 | 50 | 2/2 | 4 | 900/2000 |
T02 | 25 | 50 | 3/3 | 3 | 1800/4000 |
T03 | 75 | 25 | 2/2 | 3 | 1800/4000 |
T04 | 25 | 25 | 2/3 | 4 | 1800/4000 |
T05 | 25 | 50 | 2/2 | 3 | 1800/4000 |
T06 | 50 | 75 | 3/3 | 2 | 3600/8000 |
T07 | 25 | 50 | 3/3 | 2 | 3600/8000 |
T08 | 50 | 25 | 2/2 | 3 | 3600/8000 |
T09 | 50 | 25 | 2/2 | 2 | 3600/8000 |
T10 | 75 | 50 | 3/3 | 3 | 1800/4000 |
T11 | 75 | 25 | 2/2 | 3 | 900/2000 |
T12 | 25 | 75 | 2/2 | 3 | 1800/4000 |
N01 | 75 | 50 | 3/3 | 2 | 1800/4000 |
N02 | 75 | 75 | 3/3 | 3 | 900/2000 |
N03 | 75 | 75 | 3/3 | 4 | 900/2000 |
N04 | 75 | 25 | 3/2 | 4 | 900/2000 |
N05 | 50 | 75 | 2/2 | 4 | 3600/8000 |
N06 | 75 | 50 | 3/2 | 2 | 1800/4000 |
N07 | 50 | 25 | 3/2 | 3 | 900/2000 |
N08 | 25 | 25 | 3/3 | 3 | 3600/8000 |
N09 | 75 | 50 | 2/3 | 2 | 900/2000 |
N10 | 50 | 50 | 2/3 | 4 | 3600/8000 |
N11 | 50 | 25 | 2/2 | 3 | 3600/8000 |
N12 | 50 | 50 | 2/2 | 4 | 900/2000 |
5.3 Parameter tuning
Parameter | Range value |
---|---|
z1 | {10, 9, 8, 7} |
z2 | {6,5,4} |
z3 | {3,2,1} |
T
| [0.95, 0.9999] |
5.4 MIP tests
#Req | MIPSol | %OptGap | LB | ALNS | %Gap | |
---|---|---|---|---|---|---|
TS00 | 9/9 | 5341.04 | 12.8% | 4657.39 | 4752.56 | 2.0% |
TS01 | 7/9 | 5952.54 | 13.1% | 5175.14 | 5336.05 | 3.1% |
TS02 | 9/7 | 4699.87 | 0.0% | 4699.87 | 4699.87 | 0.0% |
TS03 | 9/7 | 5915.72 | 0.0% | 5915.72 | 5915.72 | 0.0% |
TS04 | 7/5 | 3489.58 | 0.0% | 3489.58 | 3489.58 | 0.0% |
TS05 | 9/9 | 3839.40 | 18.8% | 3117.59 | 3196.39 | 2.5% |
TS06 | 7/9 | 3335.60 | 16.2% | 2795.23 | 2877.05 | 2.9% |
TS07 | 7/9 | 5363.90 | 23.5% | 4103.38 | 4163.10 | 1.5% |
TS08 | 5/9 | 2298.82 | 4.5% | 2195.37 | 2218.41 | 1.0% |
TS09 | 5/7 | 3018.70 | 0.0% | 3018.70 | 3018.70 | 0.0% |
NS00 | 7/7 | 4127.63 | 0.0% | 4127.63 | 4127.63 | 0.0% |
NS01 | 9/7 | 4671.66 | 8.1% | 4293.26 | 4296.64 | 0.1% |
NS02 | 9/9 | 4637.60 | 8.6% | 4238.77 | 4326.73 | 2.1% |
NS03 | 9/7 | 5090.58 | 8.2% | 4673.15 | 4741.16 | 1.5% |
NS04 | 5/5 | 2922.36 | 0.0% | 2922.36 | 2922.36 | 0.0% |
NS05 | 9/9 | 4676.25 | 5.1% | 4439.63 | 4484.16 | 1.0% |
NS06 | 5/7 | 4694.72 | 4.1% | 4504.11 | 4549.69 | 1.0% |
NS07 | 5/7 | 3065.51 | 2.2% | 2998.07 | 3065.51 | 2.2% |
NS08 | 5/5 | 3770.51 | 0.0% | 3770.51 | 3770.51 | 0.0% |
NS09 | 5/7 | 3881.00 | 5.9% | 3652.02 | 3664.38 | 0.3% |
5.5 Experimental results
500–1500 | 1500–3000 | |
---|---|---|
T
| 14.86 | 2.41 |
N
| 3.81 | 0.35 |
500 iter | 1500 iter | 3000 iter | |||||||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
totCost | totDist | lhVeh | shVeh | cpu(s) | totCost | totDist | lhVeh | shVeh | cpu | totCost | totDist | lhVeh | shVeh | cpu | |
T01 | 20510.79 | 6844.12 | 9.33 | 27.33 | 27 | 19779.37 | 6446.03 | 8.33 | 26.67 | 79 | 18072.18 | 6238.85 | 7.00 | 23.67 | 455 |
T02 | 9721.03 | 3554.36 | 3 | 12.33 | 30 | 8957.17 | 3290.51 | 3.00 | 11.33 | 95 | 8938.48 | 3271.81 | 3.00 | 11.33 | 185 |
T03 | 12051.27 | 4717.94 | 4.67 | 14.67 | 16 | 11344.59 | 4511.26 | 5.00 | 13.67 | 56 | 11388.09 | 4388.09 | 4.67 | 14.00 | 113 |
T04 | 9820.73 | 3654.07 | 5 | 12.33 | 5 | 9095.25 | 3595.25 | 5.33 | 11.00 | 19 | 8625.37 | 3458.71 | 4.67 | 10.33 | 61 |
T05 | 12425.67 | 4092.34 | 3.67 | 16.67 | 21 | 11893.10 | 4059.77 | 3.33 | 15.67 | 51 | 11726.58 | 3893.25 | 3.00 | 15.67 | 136 |
T06 | 12881.36 | 4548.02 | 2.33 | 16.67 | 116 | 12684.78 | 4518.11 | 2.33 | 16.33 | 334 | 12680.61 | 4513.94 | 2.67 | 16.33 | 802 |
T07 | 9454.35 | 3621.01 | 3.33 | 11.67 | 26 | 8784.60 | 3117.93 | 2.00 | 11.33 | 78 | 8840.46 | 3173.80 | 2.00 | 11.33 | 213 |
T08 | 9262.36 | 3762.36 | 4.33 | 11 | 14 | 9018.51 | 3518.51 | 4.00 | 11.00 | 37 | 9018.09 | 3684.76 | 4.33 | 10.67 | 49 |
T09 | 7312.97 | 2979.64 | 2 | 8.67 | 11 | 7000.82 | 2834.15 | 2.00 | 8.33 | 34 | 6833.32 | 2833.32 | 2.00 | 8.00 | 69 |
T10 | 17784.05 | 6284.05 | 7.67 | 23 | 35 | 15912.66 | 5745.99 | 5.67 | 20.33 | 204 | 15492.85 | 5659.51 | 5.33 | 19.67 | 361 |
T11 | 19826.98 | 6160.31 | 8.33 | 27.33 | 21 | 19484.60 | 5984.60 | 8.67 | 27.00 | 57 | 19312.62 | 5645.95 | 8.00 | 27.33 | 105 |
T12 | 14508.28 | 4841.62 | 4.33 | 19.33 | 47 | 13258.25 | 4758.25 | 4.00 | 17.00 | 276 | 13264.96 | 4431.62 | 4.00 | 17.67 | 469 |
N01 | 11742.47 | 3909.14 | 4.67 | 15.67 | 223 | 11797.69 | 3797.69 | 4.00 | 16.00 | 605 | 11609.95 | 3776.61 | 4.33 | 15.67 | 1131 |
N02 | 19493.80 | 6827.14 | 9.67 | 25.33 | 203 | 19180.04 | 6680.04 | 9.67 | 25.00 | 1344 | 18917.06 | 6583.73 | 10.00 | 24.67 | 3813 |
N03 | 20113.55 | 7280.22 | 12.67 | 25.67 | 189 | 18793.12 | 6959.79 | 12.33 | 23.67 | 1368 | 18116.34 | 6616.34 | 9.33 | 23.00 | 2767 |
N04 | 18233.43 | 6233.43 | 10 | 24 | 19 | 16880.34 | 5713.67 | 8.67 | 22.33 | 106 | 17119.52 | 5786.19 | 9.33 | 22.67 | 184 |
N05 | 7386.80 | 3386.80 | 5 | 8 | 557 | 7352.54 | 3352.54 | 5.00 | 8.00 | 1358 | 7327.85 | 3327.85 | 5.00 | 8.00 | 2990 |
N06 | 10558.65 | 3558.65 | 4.33 | 14 | 180 | 10354.32 | 3354.32 | 4.00 | 14.00 | 446 | 10276.74 | 3276.74 | 4.00 | 14.00 | 1576 |
N07 | 13647.31 | 4813.98 | 7.67 | 17.67 | 13 | 12598.68 | 4432.02 | 7.33 | 16.33 | 66 | 12242.38 | 4075.72 | 6.33 | 16.33 | 191 |
N08 | 5218.46 | 2218.46 | 4 | 6 | 21 | 5169.43 | 2169.43 | 4.00 | 6.00 | 43 | 5174.43 | 2174.43 | 4.00 | 6.00 | 120 |
N09 | 20560.64 | 6393.97 | 8 | 28.33 | 134 | 20259.34 | 6259.34 | 8.33 | 28.00 | 407 | 20329.00 | 6329.00 | 8.67 | 28.00 | 1411 |
N10 | 7835.87 | 3669.20 | 6 | 8.33 | 112 | 7543.43 | 3543.43 | 6.00 | 8.00 | 461 | 7524.39 | 3524.39 | 6.00 | 8.00 | 1568 |
N11 | 5415.58 | 2415.58 | 4 | 6 | 51 | 5424.67 | 2424.67 | 4.00 | 6.00 | 133 | 5393.41 | 2393.41 | 4.00 | 6.00 | 501 |
N12 | 15057.63 | 5890.97 | 10.67 | 18.333 | 62 | 14258.97 | 5092.30 | 8.00 | 18.33 | 262 | 13974.41 | 5141.08 | 8.33 | 17.67 | 1523 |
Avg T | 12963.32 | 4588.32 | 4.83 | 16.75 | 30.86 | 12267.81 | 4365.03 | 4.47 | 15.81 | 110.19 | 12016.13 | 4266.13 | 4.22 | 15.50 | 251.72 |
Avg N | 12938.68 | 4716.46 | 7.22 | 16.44 | 146.99 | 12467.71 | 4481.60 | 6.78 | 15.97 | 550.09 | 12333.79 | 4417.12 | 6.61 | 15.83 | 1481.33 |
5.6 Integrated versus sequential ALNS
Sequential ALNS high-low | Sequential ALNS low-high | |||
---|---|---|---|---|
totCost | %Gap | totCost | %Gap | |
T01 | 20701.49 | 4.66% | 20625.99 | 4.28% |
T02 | 9299.62 | 3.82% | 9328.92 | 4.15% |
T03 | 11755.35 | 3.62% | 11744.46 | 3.52% |
T04 | 9408.19 | 3.44% | 9515.81 | 4.62% |
T05 | 12386.38 | 4.15% | 12188.26 | 2.48% |
T06 | 13368.35 | 5.39% | 13375.55 | 5.45% |
T07 | 9055.37 | 3.08% | 9256.22 | 5.37% |
T08 | 9512.76 | 5.48% | 9437.18 | 4.64% |
T09 | 7218.12 | 3.10% | 7237.97 | 3.39% |
T10 | 16489.77 | 3.63% | 16633.39 | 4.53% |
T11 | 19927.84 | 2.27% | 20358.03 | 4.48% |
T12 | 13931.08 | 5.07% | 13656.46 | 3.00% |
Avg | 3.96% | 4.16% | ||
N01 | 12059.05 | 2.22% | 11961.04 | 1.38% |
N02 | 19490.49 | 1.62% | 19641.89 | 2.41% |
N03 | 19224.32 | 2.29% | 19229.47 | 2.32% |
N04 | 17142.68 | 1.55% | 17393.42 | 3.04% |
N05 | 7483.01 | 1.77% | 7520.79 | 2.29% |
N06 | 10545.31 | 1.84% | 10540.53 | 1.80% |
N07 | 12883.76 | 2.26% | 12855.45 | 2.04% |
N08 | 5273.21 | 2.01% | 5288.67 | 2.31% |
N09 | 20557.74 | 1.47% | 20837.07 | 2.85% |
N10 | 7715.70 | 2.28% | 7684.95 | 1.88% |
N11 | 5525.45 | 1.86% | 5532.99 | 2.00% |
N12 | 14545.67 | 2.01% | 14662.88 | 2.83% |
Avg | 1.93% | 2.26% |